Arxiv:1905.02772V2 [Math.QA]

arXiv:1905.02772v2 [math.QA] 20 Jan 2021 φ (1.1) where aiyo oso aiod endb h eolocus qua zero the the whose study by words, defined we other manifolds particular, Poisson in In of or family construction, coincide. mirror quantisation the deformation of side either on fec polynomial each of curve lmnso h ai fgoa etosuioms uhaspectrum a such uniformise sections global a functions. on of theta structure basis algebra the explicit of an elements of spectrum the as constructed and ie ygoercqatsto ihnteSZfraim[1,wieteB the while [51], quantisation. formalism deformation SYZ by the quantised within naturally quantisation geometric by tised irrsmer.Tewr yGosHcigadKe 2]associa ( [21] pairs Keel i.e. a and A-side, its Gross-Hacking the to on by due pairs work interest jenga The enormous sparked symmetry. has mirror geometry algebraic tum bundle rjciespace projective rjciecurve projective ez ufc.I te od,tepi ( pair the words, other In surface. Pezzo nnt,i ojnapi and pair Loojenga a is infinity, ez ufc a ecniee sa ffiecn pc( Spec cone affine an as considered be can surface Pezzo ∈ rma ler-emti on fve udrcrancniin o conditions certain (under view of point algebro-geometric an From neetnl,teAsd seupe ihasmlci tutr,a structure, symplectic a with equipped is A-side the Interestingly, nti ae esuyacrancaso e ez ufcsta c that surfaces Pezzo del of class certain a study we paper this In nrcn er,suyn o-omttv ig hog h me the through rings non-commutative studying years, recent In C D [ P x φ UNIE PAINLEV QUANTISED sa nicnnclcceo ainlcre,amro aiyo h B the on family mirror a curves, rational of cycle anti-canonical an is L M 1 i tno-bokvRisqatmdlPzoadteqatmm quantum the Painlevé equations. the and of Pezzo E manifolds algebra, del Sklyanin quantum generalised the Etingof-Oblomkov-Rains T generalised cases limiting properties. the as PBW/PHS/Koszul introduce tains its we characterise particular and In algebra class. certain a Abstract. ( x , x = φ ENDCEHV AT AZCO LDMRRUBTSOV VLADIMIR MAZZOCCO, MARTA CHEKHOV, LEONID i 2 for ) M x , h rjciiainof projectivisation the KYNNADCLB-A ALGEBRAS CALABI-YAU AND SKLYANIN 3 φ fteform the of ] φ { \ WP P i ( M 1 = x nti ae esuyqatmdlPzosrae eogn t belonging surfaces Pezzo del quantum study we paper this In 0 1 φ } 3 x , φ i , , hs eto igis ring section whose ftean surface affine the of ⊂ 2 2 i , x , saplnma fdegree of polynomial a is 3 P 1 = 3 2 = ) orsodn otedivisor the to corresponding , 2 , M x 1. )tepoetv completion projective the 3) 1 oBrsDubrovin. Boris to x φ M Introduction 2 OORM MANIFOLDS, MONODROMY E = x ´ ,D Y, 3 φ M . + 1 φ φ where ) 1 M \ M ⊕ ( D x k φ φ 1 ∞ ≥ + ) D , ⊂ 0 ttesm ie ahan del affine each time, same the At . H C ∞ Y φ 0 3 ,where ), 2 ( sasot rjciesurface projective smooth a is ≤ P ( sa(osbydgnrt)del degenerate) (possibly a is M x M 2 d φ + ) C φ igfGnbr and tingof-Ginzburg ntevariable the in D [ fadegree a of L , x Sklyanin-Painlevé ∞ φ 1 ⊗ i ler con- algebra his x , D 3 ihtetiilline trivial the with , M k ( ∞ x .Hr,w althe call we Here, ). 2 x , 3 φ etrsae The space. vector ) onodromy stedvsrat divisor the is 3 nteweighted the in ] hd fquan- of thods emti and geometric / n r called are and piain in pplications tsto fa of ntisation di squan- is it nd d h h degrees the n φ e oLooi- to tes polynomial i nb put be an x vrthe over ) o i only. ieis side -side 2 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

A quantisation of a del Pezzo surface of this type appeared in the work of Oblomkov [32] as the spherical sub-algebra of the CCˇ 1 double affine Hecke algebra (DAHA). Then Etingof, Oblomkov and Rains proposed a notion of generalised DAHA for every simply laced affine Dynkin diagram and showed that their spherical sub-algebras quantise the coordinate rings of affine surfaces obtained by removing a nodal P1 from a weighted projective del Pezzo surface of degrees 3, 2 and 1 re- (1) (1) (1) spectively for E6 , E7 and E8 or by removing a triangle from a projective del (1) Pezzo surface of degree 3 in the case D4 . In the same paper, the authors defined a holomorphic (but not algebraic) map from the mini-versal deformation of the corresponding Kleinian singularity SL(2, C)/Γ (where Γ SL(2, C) is the finite ∈ subgroup corresponding to the Dynkin diagram D4, E6, E7 and E8 respectively via the McKay correspondence) to the family of surfaces where φ is in our form: Mφ (1) 2 2 2 D4 x1x2x3 + x1 + x2 + x3 + ηx1 + σx2 + ρx3 + ω, (1) 3 3 2 2 2 E6 x1x2x3 + x1 + x2 + x3 + η2x1 + η1x1 + σ2x2 + σ1x2 + ρx3 + ω, (1.2) E(1) x x x + x4 + x2 + x2 + η x3 + + η x + σx + ρx + ω, 7 1 2 3 1 3 3 3 1 ··· 1 1 2 3 E(1) x x x + x5 + x2 + x2 + η x4 + + η x + σx + ρx + ω. 8 1 2 3 1 3 3 4 1 ··· 1 1 2 3 Following this work, P. Etingof and V. Ginzburg [15] have proposed a quantum description of del Pezzo surfaces based on the flat deformation of cubic affine cone surfaces with an isolated elliptic singularity of type E6, E7 and E8 in (weighted) projective planes: x3 x3 x3 e e e E τx x x + 1 + 2 + 3 + η x2 + η x + 6 1 2 3 3 3 3 2 1 1 1 + σ x2 + σ x + ρ x2 + ρ x + ω, e 2 2 1 2 2 3 1 3 x4 x4 x2 E τx x x + 1 + 2 + 3 + η x3 + + η x + (1.3) 7 1 2 3 4 4 2 3 1 ··· 1 1 + σ x3 + + σ x + ρ x2 + ρ x + ω, e 3 2 ··· 1 2 2 3 1 3 x6 x3 x2 E τx x x + 1 + 2 + 3 + η x5 + + η x2 + η x + 8 1 2 3 6 3 2 5 1 ··· 2 1 1 1 + σ x2 + σ x + ρ x2 + ρ x + ω. e 2 2 1 2 2 3 1 3 Their result gives a family of Calabi-Yau algebras parametrised by a complex number and a triple of polynomials of specifically chosen degrees. Interestingly, as far as we know, nobody has proved a similar result for the polynomials (1.2). Poisson manifolds defined by the zero locus of a degree 3 polynomial φ Mφ ∈ C[x1, x2, x3] of the form (1.1) where φi(xi) for i =1, 2, 3 is a polynomial of degree 2 appear in the theory of the Painlevédifferential equations as monodromy manifolds [54]. Indeed, the Painlevésixth monodromy manifold is precisely the affine surface that appeared in Oblomkov [32] (see also [16]) as the spectrum of the center of the Cherednik algebra of type CCˇ 1 for q = 1. This result was generalised in [27], where seven new algebras were produced as Whittaker degenerations of the Cherednik algebra of type CCˇ 1 in such a way that their spherical–sub-algebras tend in the semi-classical limit to the monodromy manifolds of the respective Painlevédifferential equations. In the present paper we give a quantisation of the Painlevémonodromy manifolds that fits into the scheme proposed by Etingof and Ginzburg for the elliptic QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 3 singularity E6 (see Theorem 1.5). Namely, for an appropriate quantisation Φ of φ, we define an associative algebra AΦ, which is a flat deformation of the coordinate ring C[x1, x2e, x3] or, more precisely, the quantisation of the corresponding Poisson algebra A = (C[x , x , x ], , ) where φ 1 2 3 {· ·}φ dp dq dφ p, q = ∧ ∧ { }φ dx dx dx 1 ∧ 2 ∧ 3 is the Poisson-Nambu structure (2.11) on C3 for p, q C[x , x , x ]. ∈ 1 2 3 The algebra AΦ has three non-commuting generators Xi,i = 1, 2, 3 subject to the relations X X qX X = φ (X ), (i, j, k)=(1, 2, 3) i j − j i k k with φk C[Xk] and q C∗. One can consider the following diagram where the left and right∈ column arrows∈ are natural surjections and the horizontal arrows denote flat deformations or quantisations of the corresponding Poisson algebras Aφ and A / φ : φ h i /o/o fl. def./o/o /o/o/o / q (1.4) Aφ / AΦ

fl. def. q A / φ /o /o /o / A / Ω Ω0 . φ h i Φ h − i Following the idea of [15], we construct the bottom-right corner algebra as a quotient q 0 of the (family of) associative algebras AΦ by the bilateral ideal Ω Ω generated q h − i by evaluating a central element Ω AΦ for all φ corresponding to the Painlevémon- odromy manifolds. As a result, we∈ obtain a (family of) non-commutative 3-Calabi- Yau algebras that we denote by and their non-commutative 2-dimensional quo- tients as a quantum del Pezzo surfaces.UZ More precisely we give the following: m Definition 1.1. Given any scalars ǫ1,ǫ2,ǫ3, and q, q = 1 for any integer m, the universal Painlevéalgebra is the non-commutative6 algebra with generators UP X1,X2,X3, Ω1, Ω2, Ω3 defined by the relations: 1/2 1/2 1 1/2 1/2 q− X X q X X (q− q)ǫ X + (q− q )Ω =0, 1 2 − 2 1 − − 3 3 − 3 1/2 1/2 1 1/2 1/2 (1.5) q− X X q X X (q− q)ǫ X + (q− q )Ω =0, 2 3 − 3 2 − − 1 1 − 1 1/2 1/2 1 1/2 1/2 q− X X q X X (q− q)ǫ X + (q− q )Ω =0, 3 1 − 1 3 − − 2 2 − 2 [Ω , ]=0, i =1, 2, 3. i · Remark 1.2. The name universal has been chosen because in the case ǫ1 = ǫ2 = ǫ3 = 1, this algebra corresponds to the Universal Askey-Wilson algebra [52]. 0 Definition 1.3. For any choice of three scalars Ωi , i = 1, 2, 3, the confluent Zhedanov algebra is the quotient UZ / Ω Ω0, Ω Ω0, Ω Ω0 . UP h 1 − 1 2 − 2 3 − 3i Remark 1.4. The name confluent Zhedanov has been chosen because for different choices of the scalars ǫ1,ǫ2,ǫ3, the algebra is coincides with the confluent Zhedanov algebras studied in [27]. UZ Theorem 1.5. The confluent Zhedanov algebra satisfies the following properties: UZ 4 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

(1) It is a Poincar´e-Birkhoﬀ-Witt (PBW) type deformation of the homogeneous quadratic C-algebra with three generators X1,X2,X3 and the relations: 1/2 1/2 q− X1X2 q X2X1 =0, 1/2 − 1/2 (1.6) q− X X q X X =0, 2 3 − 3 2 1/2 1/2 q− X X q X X =0. 3 1 − 1 3 (2) It is a family of 3-Calabi-Yau algebras with potential

2 q 1 2 2 2 Φ := X1X2X3 qX2X1X3 + − (ǫ1X1 + ǫ2X2 + ǫ3X3 )+ UZ − 2√q (1.7) +(1 q)(Ω X +Ω X +Ω X ). − 1 1 2 2 3 3 (3) Its center Z( ) is generated by (1.8) UZ 2 ǫ2 2 2 Ω2 Ω4 := √qX3X2X1 qǫ1X X qǫ3X + √qΩ1X1 + X2 + √qΩ3X3. − 1 − q 2 − 3 √q The proof of this theorem is obtained by the combining Propositions 3.6, 4.1 and 0 4.2. The construction of the quotient / Ω4 Ω4 within the Etingof-Ginzburg framework is carried out in Theorem 4.3.UZ h − i Our quantisation is compatible with the Whittaker degeneration of generalised DAHA proposed in [27] - see Theorem 2.2 here below. In particular, we show that the Kleinian case D4 arises as a limit of the elliptic singularity case E6 - all other Kleinian cases follow as special limits as well as shown in [10]. Inspired by this, we study a broad class of degenerations of Poisson algebras in termes of rational degenerations of elliptic curves. Moreover, we connect with the work of Gross, Hacking and Keel [21], namely for each φ in the form (1.1) we produce a Looijenga pair (Y,D) where Y is the 3 smooth weighted projective completion of our affine surface φ C and D is some reduced effective normal crossing anticanonical divisorM on Y⊂ given by the divisor at infinity D . This is equipped with a symplectic structure obtained by taking the Poincaréresidue∞ of the global 3-form in weighted projective space WP3 along the divisor D . This form is symplectic on Y D = φ - this gives rise ∞ \ ∞ M to the Nambu bracket on φ. At the same time, the coordinate ring of each affine M 0 k del Pezzo φ can be seen as the graded ring k 0H (P φ,L⊗ ), where P φ M ⊕ ≥ M M is the projectivisation of φ, and L is a line bundle of an appropriate degree, defined by the anticanonicalM divisor so that the equation φ = 0 can be seen as a relation between some analogues of theta-functions related to toric mirror data on log-Calabi-Yau surfaces. Due to the fact that the Calabi Yau algebra associated to E6 specialises to the Sklyanin algebra with three generators (4.44), we provide a unified Jacobian algebra, that we call generalised Sklyanin-Painlevéalgebra, whiche for different values of the parameters specialises to the generalised Sklyanin algebra (4.45) of Iyudu and Shkarin, or to the E6-Calabi-Yau algebra of Etingof and Ginzburg or to our algebra . This is a first step towards the more ambitious goal of describing all UZ GHK theta functions ine the non-commutative world.

Definition 1.6. For any choice of the scalars a,b,c,α,β,γ,a1,b1,c1,a2,b2,c2 C, such that a,b,c are not roots of unity, the generalised Sklyanin-Painlevéalgebra∈ is QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 5 the non-commutative algebra with generators X1,X2,X3 defined by the relations:

X X aX X αX2 + a X + a =0, 2 3 − 3 2 − 1 1 1 2 (1.9) X X bX X βX2 + b X + b =0, 3 1 − 1 3 − 2 1 2 2 X X cX X γX2 + c X + c =0. 1 2 − 2 1 − 3 1 3 2 We fully characterise for which cases the generalised Sklyanin-Painlevéalgebra is a Calabi Yau algebra with PoincaréBirkhoff Witt (PBW) or Koszul properties or with a polynomial growth Hilbert series (PHS):

Theorem 1.7. For speciﬁc choices of the parameters as follows: (1) a = b = c =0 and (a3,αβγ) = ( 1, 1), (2) (a,b,c) =6 (0, 0, 0) and either6 α =− β = a b = 0 or γ = α = c a = 0 or β = γ =6 b c =0, − − (3) α = β = γ−=0 and (a,b,c) = (0, 0, 0), 6 the generalised Sklyanin-Painlev´ealgebra is potential, PHS and Koszul.

Finally, in Theorem 6.3, we deal with the question by P. Bousseau whether his deformation quantisation of function algebras on certain aﬃne varieties related to Looijenga pairs, proposed in the recent paper [6], can be compared to Etingof and Ginzburg approach.

This paper is organised as follows. In Section 2 provide some background on the Painlevémonodromy manifolds and produce their quantisation in Theorem 1.5. In particular we introduce the family of non-commutative algebras as UZ the algebra generated by X1,X2,X3 and with relations (1.5). In Section 3 we discuss the notions of PBW,h PHS andi Koszul property and show in what way the algebra satisfies them. In Section 4, we discuss the notions of Calabi Yau algebra,UZ the Etingof and Ginzburg construction and the Sklyanin algebra. We introduce the generalised Sklyanin-Painlevéalgebra (see subsection 4.8) and characterise its PBW/PHS/Koszul properties. In Section 5, we discuss the affine del Pezzo surfaces φ for different choices of φ and their degenerations in terms of rational degenerationsM of elliptic curves. In Section 6 we provide the quantum version of such elliptic degenerations. Finally in Section 7 we provide several tables that resume all these results and discuss some open questions. Acknowledgements. The authors are grateful to Yu. Berest, R. Berger, F. Eshmatov, P. Etingof, D. Gurevich, N. Iyudu, T. Kelly, T. Koornwinder, M. Gross, B. Pym, V. Sokolov, P. Terwilliger, A. Zhedanov for helpful discussions. Our special thanks to Geoffrey Powell who carefully read our first version and made several useful remarks and to the referee whose remarks were truly helpful. This research was supported by the EPSRC Research Grant EP/P 021913/1, by the Hausdorff Institute, by ANR DIADEMS and MPIM (Bonn) and SISSA (Trieste). V.R. was partly supported by the project IPaDEGAN (H2020-MSCA-RISE-2017), Grant Number 778010, and by the Russian Foundation for Basic Research under the Grants RFBR 18-01-00461 and 16-51-53034- 716 GFEN. Boris, your strength, energy, optimism and courage during the last months of your life are a testament to the great man you are. Goodbye dear friend, teacher. 6 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

2. Painleve´ monodromy manifolds and their algebraic quantisation The Painlevédifferential equations are nonlinear second order ordinary differential equations of the type: y = (t,y,y ), tt R t where is rational in y, t and yt, such that the general solution y(t; c1,c2) satisfies the followingR two important properties (see [39]): (1) Painlevéproperty: The solutions have no movable critical points, i.e. the lo- cations of multi-valued singularities of any of the solutions are independent of the particular solution chosen. (2) Irreducibility: For generic values of the integration constants c1,c2, the solution y(t; c1,c2) cannot be expressed via elementary or classical tran- scendental functions. The Painlevédifferential equations possess many beautiful properties, for example they are “integrable”, i.e. they can be written as the compatibility condition ∂A ∂B (2.10) = [B, A], ∂t − ∂λ ∂Y between an auxiliary 2 2 linear system ∂λ = A(λ; t)Y and an associated deformation system, under the× condition that the monodromy data of the auxiliary system are constant under deformation. Moreover the Painlevédifferential equations admit symmetries under affine Weyl groups which are related to the associated Bäcklund transformations. Taking these into account, to each Painlevédifferential equation corresponds a monodromy manifold, i.e. the set of monodromy data up to global conjugation and affine Weyl group symmetries. The so-called Riemann–Hilbert correspondence associates to each solution of a Painlevédifferential equation (up to Bäcklund transformations) a point in its monodromy manifold. Each monodromy manifold is an affine cubic surface in C3 defined by the zero locus of the corresponding polynomial in C[x1, x2, x3] given in Table 1, where ω1,...,ω4 are some constants (algebraically dependent in all cases except PVI) related to the parameters appearing in the corresponding Painlevé equation.

P-eqs Polynomials P V I x x x x2 x2 x2 + ω x + ω x + ω x + ω 1 2 3 − 1 − 2 − 3 1 1 2 2 3 3 4 P V x x x x2 x2 + ω x + ω x + ω x + ω 1 2 3 − 1 − 2 1 1 2 2 3 3 4 P V x x x x2 x2 + ω x + ω x + ω deg 1 2 3 − 1 − 2 1 1 2 2 4 P IV x x x x2 + ω x + ω x + ω x + ω 1 2 3 − 1 1 1 2 2 3 3 4 PIIID6 x x x x2 x2 + ω x + ω x + ω 1 2 3 − 1 − 2 1 1 2 2 4 PIIID7 x x x x2 x2 + ω x x 1 2 3 − 1 − 2 1 1 − 2 PIIID8 x x x x2 x2 x 1 2 3 − 1 − 2 − 2 P IIJM x x x x + ω x x + ω 1 2 3 − 1 2 2 − 3 4 P IIFN x x x x2 + ω x x 1 1 2 3 − 1 1 1 − 2 − P I x x x x x +1 1 2 3 − 1 − 2 Table 1. Painlev´emonodromy manifolds QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 7

Note that in Table 1, we distinguish ten different monodromy manifolds, the PIIID6 , PIIID7 and PIIID8 correspond to the three different cases of the third Painlevéequation according to Sakai’s classification [48], and the two monodromy manifolds P IIFN and P IIJM associated to the second Painlevéequation correspond to the two different isomonodromy problems found by Flaschka–Newell [17] and Jimbo–Miwa [24] respectively. Each cubic surface φ := Spec(C[x1, x2, x3]/ φ = 0 ), is endowed with the natural Poisson bracketM defined by: h i ∂φ ∂φ ∂φ (2.11) x1, x2 = , x2, x3 = , x3, x1 = . { } ∂x3 { } ∂x1 { } ∂x2 In the case of PVI, this Poisson bracket is induced by the Goldman bracket on the SL2(C) character variety of a 4 holed Riemann sphere, or is given by the Chekhov–Fock Poisson bracket on the complexified Thurston shear coordinates. In [10], all cubic surfaces were parameterised in terms of Thurston shear coordinates s1,s2,s3 and parameters p1,p2,p3 such that the Poisson bracket (2.11) is induced by the following flat one: (2.12) s ,s = s ,s = s ,s =1, 1 2 2 3 3 1 for PVI,PV,PV ,PIV,PII,PI, {p , }= p{ , =} p {, =0}, deg { 1 ·} { 2 ·} { 3 ·} s ,s = p ,s = s ,s = p ,s =1, 1 2 2 1 3 2 2 3 for PIIID6 ,PIIID7 ,PIIID8 . {s ,p } =2{, s ,s} ={ p ,} ={ p , }=0, { 2 2} { 1 3} { 1 ·} { 3 ·} We give this parameterisation in Table 2, where we think of all the monodromy manifolds as having the form1: (2.13) φ(d) =x x x ǫ(d)x2 ǫ(d)x2 ǫ(d)x2 +ω(d)x +ω(d)x +ω(d)x +ω(d) =0, P 1 2 3 − 1 1 − 2 2 − 3 3 1 1 2 2 3 3 4 where d is an index running on the list of the Painlevécubics PVI,PV,PVdeg,PIV , D6 D7 D8 JM FN (d) (d) PIII ,PIII , PIII ,PII ,PII , P I and the parameters ǫi , ωi , i = 1, 2, 3 are given by: (2.14) 1 for d = PVI,PV,PIIID6 , P V ,PIIID7 ,PIIID8 ,PIV,PIIFN , ǫ(d) = deg 1 0 for d = P IIJM ,PI,

1 for d = PVI,PV,PIIID6 , P V ,PIIID7 ,PIIID8 ǫ(d) = deg 2 0 for d = PIV,PIIFN ,PIIJM ,PI, 1 for d = PVI, ǫ(d) = 3 0 for d = PV,PIIID6 , P V ,PIIID7 ,PIIID8 ,PIV,PIIFN ,PIIJM ,PI. deg and (d) (d) (d) (d) (d) (d) (d) (d) (d) (d) (d) (d) ω1 = g1 g ǫ1 g2 g3 , ω2 = g2 g ǫ2 g1 g3 , − ∞ − − ∞ − (d) (d) (d) (d) (d) (d) (2.15) ω3 = g3 g ǫ3 g1 g2 , ∞ − − 2 2 2 2 (d) (d) (d) (d) (d) (d) (d) (d) (d) (d) (d) ω4 = ǫ2 ǫ3 g1 + ǫ1 ǫ3 g2 + ǫ1 ǫ2 g3 + g + ∞ (d) (d) (d)(d) (d) (d) (d) +g1 g2 g3 g 4ǫ1 ǫ2 ǫ3 , ∞ − 1 Note that in the current paper we have inverted the signs of x1,x2,x3 compared to [10]. 8 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV where g(d),g(d),g(d),g(d) are constants related to the parameters appearing in the 1 2 3 ∞ Painlev´eequations as described in Section 2 of [10] (note that in that paper capital (d) letters are used for the gi ).

P-eqs Flat coordinates p1 p1 p2 p2 p3 p3 g1 = e 2 + e− 2 , g2 = e 2 + e− 2 , g3 = e 2 + e− 2 , s +s +s + p1 + p2 + p3 s s s p1 p3 p3 g = e 1 2 3 2 2 2 + e− 1− 2− 3− 2 − 2 − 2 , ∞ p2 p3 p2 p3 p2 p3 p3 p2 s2+s3+ + s2 s3 s2 s3+ s3 s2+ P V I x1 = e 2 2 + e− − − 2 − 2 + e − 2 − 2 + g2e− − 2 + g3e 2 , p3 p1 p3 p1 p3 p1 p1 p3 s3+s1+ + s3 s1 s3 s1+ s1 s3+ x2 = e 2 2 + e− − − 2 − 2 + e − 2 − 2 + g3e− − 2 + g1e 2 , p1 p2 p1 p2 p1 p2 p2 p1 s1+s2+ + s1 s2 s1 s2+ s2 s1+ x3 = e 2 2 + e− − − 2 − 2 + e − 2 − 2 + g1e− − 2 + g2e 2 . p1 p1 p2 p2 p1 p2 s1 s2 s3 g1 = e 2 + e− 2 , g2 = e 2 + e− 2 , g3 = e− − − − 2 − 2 , g =1, p1 p2 ∞ s1 s2+ x1 = e− − 2 + g3e 2 , P V p2 p2 p1 p1 p2 s2 s2 2s1 p1 s1 s1 s2 x2 = e− − 2 + e− − − 2 − + g3e− − 2 + g1e− − − 2 − 2 , p1 p2 p1 p2 p1 p2 p2 p1 s1+s2+ + s1 s2 s1 s2+ s2 s1+ x3 = e 2 2 + e− − − 2 − 2 + e − 2 − 2 + g1e− − 2 + g2e 2 . p1 p1 p2 p2 g1 = e 2 + e− 2 , g2 = e 2 + e− 2 , g3 =0, g =1 p1 p2 p2 ∞ p2 p1 s1 s2 2s1 s2 p1 s1 s2 P Vdeg x1 = e− − 2 , x2 = e− − 2 + e− − − − 2 + g1e− − − 2 − 2 , p1 p2 p1 p2 p1 p2 p2 p1 s1+s2+ + s1 s2 s1 s2+ s2 s1+ x3 = e 2 2 + e− − − 2 − 2 + e − 2 − 2 + g1e− − 2 + g2e 2 . p1 p1 p2 p1 + s1 s2 s3 g1 = e 2 + e− 2 , g2 = e 2 , g3 =0, g = e− − − − 2 , 2s1 s2 2s3 p1 2s1 s2 s3 p1 ∞ x1 = e− − − − + e− − − − , P IV p1 2s1 s2 p1 s2 2s1 s2 s3 p1 s1 s2 x2 = e− − − + e− + e− − − − + g1e− − − 2 , p1 s3 s1+ x3 = e− + g2e 2 . s2 p2 s2 p2 s1+ + +s3+ g1 = g3 =1, g2 = e 2 2 , g = e 2 2 , s2 p2 s2 p2 s2 ∞p2 s2 p2 s2 p2 + s1 + +s3+ s1 +s3+ s1+ +s3+ x1 = e− 2 2 + e − 2 2 + e− 2 2 + e − 2 2 + e 2 2 , PIIID6 s1 s1 s2 s2 s3 s1+s3 s2+s3 s1 s2+s3 s1+s2+s3 x2 = e + e − e− + e +2e + e− + e − + e , s p s p s p 2 2 −2 2 2 + 2 x3 = e− 2 − 2 + e 2 − 2 + e 2 2 . s2 p2 s2 p2 s2 p2 s2 p2 s1+ + + s1 + s1+ + g1 =1, g2 = e 2 2 , x1 = e− 2 2 + e − 2 2 + e 2 2 , p2 D s2+ s1 s1 s2 s2 s1+s2 PIII 7 g3 =0, g = e 2 , x2 =1+2e + e − + e− + e , s2 p2∞ s2 p2 s2 + p2 x3 = e− 2 − 2 + e 2 − 2 + e 2 2 . s2 p2 s2 p2 s2 p2 2 + 2 2 + 2 2 + 2 D g1 =1, g2 = g = e , g3 =0, x1 = e− + e , PIII 8 ∞ s2 p2 s2 p2 s2 p2 s2 s2 + x2 =2+ e− + e , x3 = e− 2 − 2 + e 2 − 2 + e 2 2 . + p2 g1 = g3 = g =1, g2 = e 2 , P IIJM ∞ s1 s1 s3 s3 s1+s3 s2 s3 s3 x1 = e− + e− − , x2 = e + e , x3 = e− − + e− . s1 s2 s3 s2+s3 g1 = e− − − , g2 = g =1, g3 =0, x1 = e , P IIFN 2s3+s1+s2 2s3+s2 ∞ s1 s2 s2 s3 s2 s3 x2 = e + e + e− − + e− , x3 = e− + e− − . g1 = g2 = g =1, g3 =0, P I s1 ∞ s1 s2 s2 s1+s2 s1 x1 = e− , x2 = e− − + e− , x3 = e + e . Table 2. Flat coordinates on the Painlev´emonodromy manifolds QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 9

We recall that the celebrated confluence scheme of the Painlevédifferential equations is the following diagram,

D6 / D7 / D8 PIII / PIII / PIII ⑤= ❉❉ ③< ❉❉ ⑤⑤ ❉❉ ③③ ❉❉ ⑤⑤ ③❉③❉ ❉❉ ⑤ ③③ ❉❉ ❉❉ ⑤⑤ ③ " ❉" P / P / P deg P JM / P V I V ❇ V < II < I ❇❇ ❊❊ ②< ③③< ❇❇ ❊❊ ②② ③③ ❇ ②❊②❊ ③③ ❇❇ ②② ❊❊ ③③ ❇! ②② " ③ / FN PIV / PII where the arrows represent confluences, i.e. degeneration procedures where the independent variable, the dependent variable and the parameters are rescaled by suitable powers of ε and then the limit ε 0 is taken. This was studied on the level of monodromy manifolds in [10]. → Here, we provide the quantisation of all the Painlevécubics and produce the corresponding quantum confluence in such a way that quantisation and confluence commute. To produce the quantum Painlevécubics, we introduce the Hermitian operators S1,S2,S3, P1, P2, P3 subject to the commutation rule inherited from the Poisson bracket of s1,...,p3: [P , ]=0, [S ,S ]= iπ~ s ,s j =1, 2, 3, j +3 j. j · j j+1 { j j+1} ≡ Observe that the commutators [Si,Sj ] are always numbers and therefore we have ab exp(aS )exp(bS ) = exp aS + bS + [S ,S ] , i j i i 2 i j for any two constants a,b. Therefore we have the Weyl ordering: S +S 1 S S 1 S S iπ~ (2.16) e 1 2 = q 2 e 1 e 2 = q− 2 e 2 e 1 , q e− . ≡ (d) (d) After quantisation, the parameters g1 ,...,g that are not equal to 0 or 1 become (d) (d) ∞ Hermitian operators G1 ,...,G and are automatically Casimirs. We define the (d) (d) ∞ (d) operators Ωi in terms of G1 ,...,G by the same formulae (2.15) that link the (d) (d) ∞ (d) ωi to the gi ’s - these are also Casimirs. The parameters ǫi are scalars, and they remain scalar under quantisation. We introduce the Hermitian operators X1,X2,X3 as follows: consider the classical expressions for x1, x2, x3 is terms of s1,s2,s3 and p1,p2,p3 as in Table 2 in which each additive term is written as exponential of a sum of integer linear combinations of s1,s2,s3 and p1,p2,p3 . Replace those exponents by their quantum version. For example the quantum version of x1 in the P V I case is

P2 P3 P2 P3 P2 P3 S2+S3+ + S2 S3 S2 S3+ X1 = e 2 2 + e− − − 2 − 2 + e − 2 − 2 + P2 P2 S P3 P3 P3 S + P2 +(e 2 + e− 2 )e− 3− 2 + (e 2 + e− 2 )e 2 2 .

In this way, the resulting operators are Hermitian, because S1,S2,S3, P1, P2, P3 are. Then, the following result establishes a relation between the quantisation of the Painlevémonodromy manifolds and the confluent Zhedanov algebra given in Definition 1.1: 10 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

(d) (d) (d) Proposition 2.1. The Hermitian operators X1,X2,X3, Ω1 , Ω2 , Ω3 generate the algebra C X ,X ,X , Ω(d), Ω(d), Ω(d) / J ,J ,J ,J with h 1 2 3 1 2 3 i h 1 2 3 4i 1/2 1/2 1 (d) 1/2 1/2 (d) J = q− X X q X X (q− q)ǫ X + (q− q )Ω , 1 1 2 − 2 1 − − 3 3 − 3 1/2 1/2 1 (d) 1/2 1/2 (d) (2.17) J = q− X X q X X (q− q)ǫ X + (q− q )Ω , 2 2 3 − 3 2 − − 1 1 − 1 1/2 1/2 1 (d) 1/2 1/2 (d) J = q− X X q X X (q− q)ǫ X + (q− q )Ω , 3 3 1 − 1 3 − − 2 2 − 2 J = [Ω(d), ], i =1, 2, 3. 4 i · (d) where ǫi are the same as in the classical case. Proof. The proof of this result is obtained by direct computation by using the definitions of the quantum operators X1,X2 and X3 in terms of S1,S2,S3. By applying the quantum commutation relations for S1,S2,S3 (2.16), relations (1.5) follow. In [10], we showed that the confluence procedure for the Painlevé differential equations corresponds to certain limits of the shear coordinates, for example for PVI to PV is obtained by the substitution p3 p3 2 log ε in the limit ε 0. We define the quantum confluence by the same rescaling→ − the quantum Hermitian→ operators by ε and taking the same limit as ε 0. For example, by imposing exactly the same → limiting procedure on P3, we obtain a limiting procedure on the quantum operators (V I) (V I) (V I) X1,X2,X3, Ω1 , Ω2 , Ω3 satisfying relations (2.17) for d = V I, that produces (V ) (V ) (V ) some new quantum operators X1,X2,X3, Ω1 , Ω2 , Ω3 . By construction, these operators satisfy relations (2.17) for d = V . The same construction can be repeated for every d. Therefore, we have the following: Theorem 2.2. The confluence of the Painlevéequations commutes with their quantisation.

3. Poincare-Birkhoff-Witt´ (PBW)-deformation properties of the quantum algebra (1.5) In this section we study the algebraic properties of the quantum algebra . UZ The basic observation is that when all constants ǫi and all values of the Casimirs Ωi, i = 1, 2, 3, are zero, then (1.5) are standard quantum commutation relations defining a graded algebra that is a PBW deformation of the polynomial algebra in three variables. Here we adapt the work of [41] and [7] to check that is a PBW UZ type deformation for all cases of ǫi and all values of the Casimirs Ωi, i =1, 2, 3. 3.1. To PBW or not to PBW. Here we discuss the definition of the PBW, PHS and Koszul properties. Let V be a finite-dimensional K-vector space of dimension n with basis x n . { i}i=1 Consider the tensor algebra T •(V ) of V over K - this is the free associative algebra T •(V ) = K x ,...,x . For any pair of integers 1 i < k n we choose an h 1 ni ≤ ≤ element Ji,k T •(V ) such that deg Ji,k 2. Let J be the union of the bilateral ideals ∈ ≤ x x x x J i ⊗ k − k ⊗ i − i,k in T •(V ). Then the quotient algebra A = T •(V )/ J is equipped with the ascending h i filtration Fk , k 1; F 1 = 0 (i.e. Fk 1 Fk ) such that Fk consists of all elements of{ degree} ≥ −k in −x ,...,x . − ⊂ ≤ 1 n QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 11

Deﬁnition 3.1. The (ﬁltered) unital associative algebra A is said to satisfy the PBW property if there is an isomorphism of graded algebras

k 0Fk/Fk 1 S(V ), ⊕ ≥ − ≃ where S(V ) is the symmetric algebra of V . [41] Given a filtered algebra A with filtration by finite-dimensional vector spaces, we write P (A) := dim(A )tk Z[[t]] t k ∈ k Z X∈ for the Hilbert-Poincaréseries of the associated graded algebra

gr(A)= k 0Ak := k 0Fk/Fk 1. ⊕ ≥ ⊕ ≥ − For the purposes of this paper, we distinguish the case of n = 3 and give the following definition: Definition 3.2. The algebra A is said to satisfy the PHS property if its Poincaré- 1 Hilbert series of A coincides with (1 t)3 . We shall call PHS-algebras 3-algebras with this property. [22] −

In the case of a Lie algebra g of dimension n with a basis x1, ..., xn , there is a natural reformulation of the PBW-property for the universal{ enveloping} algebra U(g) in terms of the map σ : S(g) gr(U(g)) where S(g) is the symmetric algebra of the Lie algebra g and gr(U(g))→ is the associated graded algebra of the filtered algebra T •(g). This map is defined due to the universality of U(g) from the relation σ τ = φ where τ : T •(g) S(g) is the canonical projection and φ : T •(g) gr(◦U(g)) is the surjective morphism→ of graded algebras induced by the canonical→ projection of T •(g) U(g). In this case, the following→ three statements are equivalent (see [20]): the homomorphism σ : S(g) gr(U(g)) is a graded algebra isomorphism; • → if g admits a totally ordered basis xλ λ Λ then the subset • { } ∈ 1 x ...x (λ ,...,λ ) Λn, λ ... λ , n 1 { }∪{ 1 λn | 1 n ∈ 1 ≤ ≤ n ≥ } gives a basis of U(g); the canonical map g U(g) is an injection. • → We shall use these reformulations of PBW to choose among them a form which is convenient to our aims. We conclude this subsection by recalling the definition of Koszul algebra. Let A be a graded algebra over a field K of characteristic 0:

A = ∞ A , ⊕k=0 k its augmentation ideal A+ is by deﬁnition + A := ∞ A ⊕k=1 k and the canonical projection π : A A = A/A+, 7→ 0 is called augmentation map. By the augmentation map, A0 can be considered as an A-module: A A A , (a, x)= π(a)x. × 0 → 0 12 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

Definition 3.3. (Koszul Algebras). A Koszul algebra A is an N-graded algebra A = ∞ A over a field K that satisfies following conditions: ⊕k=0 k A0 = K. • + A0 A/A , considered as a graded A-module, admits a graded projective • resolution≃ P (2) P (1) P (0) A 0. ···→ → → → 0 → such that P (i) is generated as a Z-graded A-module by its degree i compo- nent, i.e., for the decomposition of A- modules: (i) (i) P = j ZPj ⊕ ∈ (i) (i) one has that P = APi . Standard examples of Koszul algebra are the symmetric algebra S(V ) and the exterior algebra Λ(V ) of an n dimensional K-vector space V .

Given a Koszul algebra A = k∞=0Ak, consider the tensor algebra T (A1) and the map ⊕ µ : T (A ) A, µ(x x ) := x ...x . 1 → 1 ⊗···⊗ k 1 k A classical theorem (see for example [42]) states that every Koszul algebra is quadratic, namely, A T (A )/ I ≃ 1 h i where I us the ideal generated by the quadratic relation: h i (ker µ) (A K A ). ∩ 1 ⊗ 1 The inverse statement is not always true. Priddy ([42]) proved that if a homogeneous quadratic algebra has a PBW basis, then it is Koszul. 3.2. PBW-type algebra structure. In this sub-section, we follow the work by Braverman-Gaitsgory [7] to adapt the ideas of sub-section 3.1 to the case of non homogeneous algebras such as our quantum algebra . The free non-commutative polynomial associativeUZ algebra C X ,X ,X can be h 1 2 3i considered as the tensor algebra T •(V ), where V = Vect X1,X2,X3 , that is filtered by the natural filtration: h i k j F (T •(V )) = j kT (V ) . {⊕ ≤ } We are now going to explain how this filtration descends to the quotient. 2 Fix a subspace Iˆ F (T •(V )) = C V (V V ) and let I V V be the ⊂ ⊕ ⊕ ⊗ 2 ⊂ ⊗ image of Iˆ : I = π(Iˆ) under the natural projection π : F (T •(V )) V V . There → ⊗ is a epimorphism of graded algebras (denoted by the same letter) π : T •(V )/ I h i → gr(T •(V )/ Iˆ ). h i Definition 3.4. [7] The non-homogeneous quadratic algebra ˆ = T •(V )/ Iˆ is a A h i PBW-type deformation of := T •(V )/ I if the projection π is an isomorphism of graded algebras. A h i

2 Remark 3.5. If the subspace Iˆ F (T •(V )) (as it is in our case) the algebra ⊂ = T (V )/ I is usually called a 2-homogeneous algebra. The deformation ˆ = A h i A T (V )/ Iˆ is deﬁned by two linear maps l : I V and l : I C in the sense that h i 1 → 2 → ˆ = T (V )/ Iˆ T (V )/(u l (u) l (u) u I). A h i≃ − 1 − 2 | ∈ QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 13

The linear maps l1,2 satisfy some Jacobian type conditions as in Theorem 3.7 here below. Roughly speaking, this means that the graded algebra gr( ˆ) associated to the A filtered non-homogeneous quadratic algebra ˆ = T •(V )/ Iˆ is the 2-homogeneous A h i quadratic = T •(V )/ I . To showA that our quantumh i algebra is a PBW-type deformation, the first step is to show that it admits a naturalUZ filtration. This is obtained by considering it as a quotient of the free polynomial associative algebra with three generators C X1,X2,X3 by non-homogeneous relations with linear and affine terms. Then, weh need to provei that π is indeed an isomorphism, namely we need to prove the first statement of Theorem 1.5: Proposition 3.6. The quantum algebra is a PBW type -deformation of the UZ homogeneous quadratic C-algebra with three generators X1,X2,X3 and the relations (1.6). Proof. The demonstration consists of two steps. First, we drop linear and constant terms and consider the “purely” quadratic algebra and show that it is a standard A PBW-deformation of the polynomial free algebra C X1,X2,X3 with three generators. By quotienting out the relations (1.6) we obtainh a graded algebi ra and one can p s r easily see (choosing, for example, the base of ordered monomials X1 X2 X3 ) that the dimension of the homogeneous components of this algebra for different s = 0 is constant (flat-deformation). 6 As second step, we consider the non homogeneous algebra ˆ (for generic q). This is based on the application of the following theorem due toA Braverman-Gaitsgory [7] and Polishchuk-Positselsky [41] to the homogeneous ideal (I) generated by the relations I V V (1.6): I ⊂ ⊗ Theorem 3.7. [7] Let ˆ be a non homogeneous quadratic algebra, ˆ = T •(V )/ Iˆ , A A h i and = T •(V )/ I ) its corresponding 2-homogeneous quadratic algebra. Suppose A h i is a Koszul algebra. Then ˆ is a PBW-type deformation of if and only if there existA linear functions l : I AV, l : I C for which A 1 → 2 → Iˆ = u l (u) l (u) u I . { − 1 − 2 | ∈ } and the following conditions are satisfied Im(l Id Id l ) I; • 1 ⊗ − ⊗ 1 ⊆ l1(l1 Id Id l1)= (l2 Id Id l2), • l (l ⊗ Id − Id ⊗ l )=0−, ⊗ − ⊗ • 2 1 ⊗ − ⊗ 1 where the maps l Id Id l and l Id Id l are defined on the subspace 1 ⊗ − ⊗ 1 2 ⊗ − ⊗ 2 (I V ) (V I) T •(V ). ⊗ ∩ ⊗ ⊂ Remark 3.8. For the case of the finite-dimensional Lie algebra g, one has ˆ = U(g) and = S(g), the symmetric algebra of g. Consider I g g definedA as A ⊂ ⊗ I = x1 x2 x2 x1, x1, x2 g . Then l1(x1 x2 x2 x1) := [x1, x2], l2 := 0. The{ three⊗ conditions− ⊗ in Theorem∈ } 3.7 are equivalent⊗ − to the⊗ Jacobi identity. Polishchshuk and Positselsky studied the conditions for PBW property for quadratic algebras in a more general setting ([41]). We shall reformulate the conditions in theorem 3.7 in a form that is easy to verify in our case (see Theorem 2.1 ch.5 in [41]) , i.e. in terms of the bracket operator [ , ]: I V V V satisfying two conditions · · ⊂ ⊗ → 14 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

(3.18) [ , ] [ , ] : (I V ) (V I) I · · 12 − · · 23 ⊗ ∩ ⊗ →

(3.19) [ , ]([ , ] [ , ] ) : (I V ) (V I) 0. · · · · 12 − · · 23 ⊗ ∩ ⊗ 7→ We remark that the subspace (I V ) (V I) V V V ⊗ ∩ ⊗ ⊂ ⊗ ⊗ defines an analog of the space of symmetric elements of degree 3. The bracket operation [ , ]: I V V V is defined only on the subspace I that is why, due to the first· condition· ⊂ ensures⊗ → that the bracket maps I V V I again into I and we can apply it once more. ⊗ ∩ ⊗ In this setting, the map l1 Id Id l1 is given by [ , ]12 [ , ]23 while the map l Id Id l is [ , ]([ , ] ⊗ [ ,−] ).⊗ · · − · · 2 ⊗ − ⊗ 2 · · · · 12 − · · 23 As mentioned before, if the quadratic algebra = T •(V )/ I is Koszul then A h i the associated graded algebra gr( ˆ) where ˆ = T •(V )/ I [ , ]I is isomorphic to gr( ). Here, I [ , ]I means theA space of elementsA u h[ ,−]u,· · u i I and the ideal A − · · − · · ∈ I [ , ]I co-incides with the non-homogeneous ideal Iˆ . h − · · i h i Let us check the conditions (3.18), (3.19) in the case when V = CX CX CX 1 ⊕ 2 ⊕ 3 and T •(V )= C X ,X ,X and Iˆ is the ideal generated by relations (1.5): h 1 2 3i h i (3.20) ˆ = T •(V )/ Iˆ . A h i The first condition (3.18) is valid straightforwardly. The second condition (3.19) follows from the following equality

(X1X2 qX2X1)X3 + (X2X3 qX3X2)X1 + (X3X1 qX1X3)X2 = (3.21) − − − X (X X qX X )+ X (X X qX X )+ X (X X qX X ), 3 1 2 − 2 1 1 2 3 − 3 2 2 3 1 − 1 3 that is proved by replacing the quadratic terms in the brackets by L1,L2,L3, where

1/2 3/2 (d) (d) L := (q− q )ǫ X (1 q)Ω , 1 − 3 3 − − 3 1/2 3/2 (d) (d) (3.22) L := (q− q )ǫ X (1 q)Ω , 2 − 1 1 − − 1 1/2 3/2 (d) (d) L := (q− q )ǫ X (1 q)Ω , 3 − 2 2 − − 2 leading to the identity

L1X3 + L2X1 + L3X2 = X3L1 + X1L2 + X2L3, that is trivially satisfied due to the fact that [Li,Xi] = 0. To conclude, the “pure quadratic” part is Koszul hence, the non-homogeneous A algebra ˆ is a flat deformation of the polynomial algebra C[X ,X ,X ]. A 1 2 3 Remark 3.9. Braverman and Gaitsgory gave a fairly simple proof that the Koszul property of and the conditions (3.18) and (3.19) i.e. PBW-property imply the A existence of a graded deformation ~ of ˆ such that at ~ = 1 it is canonically A A isomorphic to ˆ. This is what we shall understand under “good” (or “flat”) deformation properties.A QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 15

3.3. Zhedanov algebra and its degenerations. As explained in Section 2, the quantum algebras of definition 1.3 are quantisations of the monodromy manifolds of the Painlevédifferential equations. The Painlevésixth monodromy manifold appeared in the paper by Oblomkov [32] (see also [16]) as the spectrum of the center of the Cherednik algebra of type Cˇ1C1 for q = 1. This result was generalised in [9] where this affine cubic surface was explicitly quantised leading to the Zhedanov algebra, which is isomorphic [25] to the spherical sub–algebra of the Cherednik algebra of type Cˇ1C1. In [27], seven new algebras were produced as confluences of the Cherednik algebra of type Cˇ1C1 in such a way that their spherical–sub- algebras tend in the semi-classical limit to the monodromy manifolds of all other Painlevédifferential equations. The quantum algebras defined by relations (1.6) are isomorphic to the spherical–sub-algebras introduced in [27], in the same way in which the Zhedanov algebra is isomorphic to the spherical sub–algebra of the Cherednik algebra of type Cˇ1C1. Our Theorem 1.5 shows that the Zhedanov algebra and its degenerations are flat deformations of the polynomial algebra C[X1,X2,X3].

4. Relation with Calabi-Yau and Sklyanin algebras The aim of this section is to clarify the relations of our quantum algebra with the quantum analogues of del Pezzo surfaces introduced by Etingof and GinzburgUZ [15]. The latter are elements of a very general class of non-commutative algebras related to the twisted Calabi-Yau algebras introduced by V. Ginzburg [53]. We start by recalling these notions here.

4.1. Calabi-Yau algebras and potentials. Let A be a ﬁnite dimensional, associative and graded C-algebra . We say that A is d-Calabi-Yau of dimension d if d n ExtA(A, A A) A as a bimodule and otherwise (n = d)) ExtA(A, A A)=0. In this paper,⊗ we will≃ focus on the case of 3-Calabi-Yau algebras.6 Ginzburg⊗ has argued that most 3-Calabi-Yau algebras arise as a certain quotient of the free associative algebra. More precisely, let V be a C-vector space with base X1,X2,X3; its tensor algebra T •(V ) is the free associative graded algebra A := C X1,X2,X3 . One can consider the elements of C X ,X ,X as non-commutativeh words obtainedi from h 1 2 3i the variables X1,X2,X3. The quotient T •(V )/[T •(V ),T •(V )] is the space of cyclic words or “traces”. This is the 0-degree Hochschild homology of the free algebra C X ,X ,X . We shall use in what follows the usual notation for the quotient of h 1 2 3i an associative algebra by the space of commutators, A♮ := A/[A, A]. One can deﬁne cyclic derivatives ∂ ∂ for any Φ A by j ≡ Xj ∈ ♮

(4.23) ∂j Φ := Xi +1Xi +2...XiN Xi Xi ...Xi 1 A, k k 1 2 k− ∈ k i =j |Xk where j =1, 2, 3 and all indices i ,...,i (1, 2, 3). 1 N ∈ The two-sided ideal JΦ =< ∂1Φ, ∂2Φ, ∂3Φ > in A is a non-commutative analogue of the Jacobian ideal and we can pass to the quotient

(4.24) AΦ := A/JΦ.

3 We say that the an element Φ (F T •(V ))♮ is a Calabi-Yau potential if AΦ is a 3-CY-algebra. ∈ 16 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

4.2. Etingof-Ginzburg quantisation. Given a polynomial φ C[x1, x2, x3] such that the compactifying divisor of its zero set is smooth, Etingof∈ and Ginzburg constructed an associative algebra AΦ which is a flat deformation of the coordinate ring C[x1, x2, x3] or, more precisely, the quantisation of the corresponding Poisson algebra A = (C[x , x , x ], , ) where φ 1 2 3 {· ·}φ dP dQ dφ P,Q = ∧ ∧ { }φ dx dx dx 1 ∧ 2 ∧ 3 3 is the Poisson-Nambu structure (2.11) on C for P,Q C[x1, x2, x3] [15]. Let us remind that the flat deformations of a Poisson∈ algebra (A, π) are governed by the second group of Poisson cohomology HP 2(A) and a flatness of the Poisson algebra means also a flatness of a deformation of A as a commutative algebra. The flat deformations considered by Etingof and Ginzburg are semiuniversal deformations with smooth parameter scheme such that the Kodaira-Spencer map is a vector space isomorphism. As a consequence of the computations in [38] (see Proposition 3.2), the family of affine Poisson brackets (2.11) is a family of unimodular Poisson brackets, so by the result of Dolgushev ([14]) the quantisation AΦ is a Calabi-Yau algebra generated by three non-commutative generators Xi,i =1, 2, 3 subject to the relations ∂Φ ∂Φ ∂Φ = = =0, ∂X1 ∂X2 ∂X3 where Φ is a potential whose non commutative Jacobian ideal is a suitable quantum analogue of the classical Jacobian ideal in the local algebra of φ. In [15], the authors quantise the natural Poisson structure on the hyper-surface 3 in C with an isolated elliptic singularity of type E˜r, r =6, 7, 8. Such hyper-surfaces are the zero locus of the weighted homogeneous part of the polynomials (1.3) in P2, WP1,1,2 and WP1,2,3 respectively: 3 3 3 (6) x1 x2 x3 E6 φ = τx1x2x3 + + + , ∞ 3 3 3 4 4 2 (7) x1 x2 x3 (4.25) Ee7 φ = τx1x2x3 + + + , ∞ 4 4 2 6 3 2 (8) x1 x2 x3 Ee8 φ = τx1x2x3 + + + . ∞ 6 3 2

In each case, their quantisation produces a 3-Calabi-Yau algebra A (r) defined by a e Φ∞ suitable quantum potential Φ(r), r =6, 7, 8. Motivated by the study of miniversal ∞ deformations of elliptic singularities, Etingof and Ginzburg study deformations of the potential Φ(r) by adding a term of the form ∞ (4.26) Ψr = P (X1)+ Q(X2)+ R(X3), where the polynomials P,Q and R depend of a total of µ arbitrary parameters, µ being the Milnor number of the elliptic singularity, and have smaller degree than (r) Φ has in the variable X1, X2 and X3 respectively. They prove that for such ∞ (r) choice of Ψr the sum potential Φr := Φ +Ψr also defines 3-Calabi-Yau algebra ∞ AΦr with central element Ωr that in the classical limit tends to the full polynomial φr in (1.3). This central element, let us drop the index r to keep the discussion general, Ω is used in [15] as a non-commutative analogue of the polynomial φ and the QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 17 quotient AΦ/(Ω) is a non-commutative analogue of the Poisson algebra Aφ/(φ). As a consequence, the authors consider the following commutative diagram where the left and right column arrows are natural surjections and the wave-like arrows denote flat deformations (or quantisations) of the corresponding Poisson algebras Aφ and Aφ/(φ):

ﬂ./o/o/o def./o/o /o/o / q (4.27) Aφ / AΦ

fl. def. q A / φ /o /o /o / A / Ω Ω0 . φ h i Φ h − i The idea of [15] is to construct the bottom-right corner algebra as a quotient of the q 0 (family of) associative algebras AΦ by the bilateral ideal Ω Ω generated by a q h − i central element Ω Z(AΦ). At the quantum∈ level, the difficulty is that the potential Φ and the central element Ω are different, even though, in their classical limit, they produce the same q polynomial φ. As a consequence, to complete the construction of AΦ/(Ω) one needs to find the explicit expression for Ω. In [15] this was done explicitly for the elliptic singularities of type E6, E7 and E8. Let us describe the E˜6 case in some detail - as we shall see, this is the specific case that in certain singulare limite producese the monodromy manifolds of the Painlevé equations - note that the compactifying divisor of the corresponding affine del Pezzo surfaces is always given by a triangle of lines, and therefore is not smooth. For this reason, the proofs of sub-section 4.3 are not a straight consequence of the results in [15]. It is convenient to recast the polynomial φ6 in the form t 1 φτ,t = τx x x + (x3 +x3 +x3)+ (a x2 +b x2 +c x2)+a x +b x +c x +d. a,b,c,d 1 2 3 3 1 2 3 2 1 1 1 2 1 3 2 1 2 2 2 3 τ,t Let us denote by A τ,t /(φa b c ) the coordinate ring defined by the affine Poisson φa,b,c,d , , ,d τ,t C3 surface φa,b,c,d = 0 in . (d) τ,t We note that φP defined in (2.13) is a specialisation of φa,b,c,d corresponding to the choice of parameters: (4.28) τ =1, t =0, a = ( 2ǫ(d),ω(d)), b = ( 2ǫ(d),ω(d)), c = ( 2ǫ(d),ω(d)), d = ω(d), − 1 1 − 2 2 − 3 3 4 P1 P so that the bundle over the projectivisation MφP of the the surface MP coincides with our general isomonodromic cubic surface (2.13).

Etingof and Ginzburg consider the family of homogeneous potentials ΦEG C X ,X ,X ∈ h 1 2 3i♮ t (4.29) Φ = X X X qX X X (X3 + X3 + X3), EG 1 2 3 − 2 1 3 − 3 1 2 3 q and show that the ﬁltered algebra AΦEG with generators X1,X2,X3 subject to the relations X X qX X = tX2, 1 2 − 2 1 3 X X qX X = tX2, 2 3 − 3 2 1 X X qX X = tX2, 3 1 − 1 3 2 18 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV is a 3-CY algebra. They then add a deformation potential ΨEG where 1 (4.30) Ψ = (a X2 + b X2 + c X2)+ a X + b X + c X + d. EG 2 1 1 1 2 1 3 2 1 2 2 2 3 Note that this deformation potential is precisely in the form (4.26) with 1 1 1 1 1 1 P = a X2 + a X + d, Q = b X2 + b X + d, R = c X2 + c X + d. 2 1 1 2 1 3 2 1 2 2 2 3 2 1 3 2 3 3

The sum ΦEG +ΨEG depends on q and further 8 parameters a = (a1,a2), b = (b1,b2), c = (c1,c2), q and t (the Milnor number of the corresponding elliptic Goren- stein singularity is 8). The Jacobian of the potential ΦEG +ΨEG gives the following relations

X X qX X tX2 + c X + c =0, 1 2 − 2 1 − 3 1 3 2 (4.31) X X qX X tX2 + a X + a =0, 2 3 − 3 2 − 1 1 1 2 X X qX X tX2 + b X + b =0, 3 1 − 1 3 − 2 1 2 2 that deﬁne the family of algebras Aq . ΦEG+ΨEG The Theorem 3.4.4 from [15] claims that for generic values of the parameters q,a ,a ,b ,b ,c ,c ,d,t, the family of algebras Aq is Calabi-Yau. 1 2 1 2 1 2 ΦEG+ΨEG In fact, they prove a more general statement; in each case we may choose

X X X qX X X t (X3 + X3 + X3), for E 1 2 3 − 2 1 3 − 3 1 2 3 6 (4.32) Φ = X X X qX X X t 1 X4 + 1 X4 + 1 X2 , for E EG  1 2 3 2 1 3 4 1 4 2 2 3 7 X X X − qX X X − t 1 X6 + 1 X3 + 1 X2 , for E  1 2 3 − 2 1 3 − 6 1 3 2 2 3 8 and taking ΨEG =P (X1)+ Q(X2)+ R(X3) depending on generic µ+1 parameters with P , Q, R non-homogeneous polynomials of degree:

2, for E6, 2, for E6, 2, for E6, deg(P )= 3, for E , deg(Q)= 3, for E , deg(R)= 1, for E ,  7  7  7  5, for E8,  2, for E8,  1, for E8, the sum ΦEG +ΨEG is a Calabi-Yau potential and its Jacobian defines a filtered family of associative 3-Calabi-Yau algebras with µ + 1 parameters, where µ is the Milnor number of the respective Gorenstein singularity. q In each case Er, r = 6, 7, 8, this family of filtered algebras AΦEG+ΨEG forms q the Rees algebras of the corresponding algebras AΦEG with homogeneous potentials e q q ΦEG given in (4.32). The algebras AΦEG+ΨEG /(Ω) where Ω Z(AΦEG+ΨEG ) is a non-scalar central element, give a semi-universal family of assoc∈ iative algebras (depending on q and µ parameters) which are 3-Calabi-Yau as well. q The theorem 3.4.5 in [15] proves that the center Z(AΦEG+ΨEG ) is the polynomials C q q algebra [Ω] and the quotient-algebra AΦEG+ΨEG /Ω of AΦEG+ΨEG by the two-sided τ,t ideal Ω gives a flat deformation of Aφτ,t /(φa,b,c,d). h i a,b,c,d The main difficulty in this description, as it remarked by Etingof and Ginzburg, is to compute the explicit form of the Casimir Ω. In the case of E6, the central element is given by [15, 43]: e QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 19

(4.33) Ω = ( a2q2 a qt 2a q2t a q3t b c qt2)X + EG − 1 − 2 − 2 − 2 − 1 1 1 t( b 2b q 2b q2 b q3 a c qt + b2t2 b t3 b qt3)X + − 2 − 2 − 2 − 2 − 1 1 1 − 2 − 2 2 t( c q 2c q2 2c q3 c q4 a b qt c2qt2 + c t3 + c qt3)X + − 2 − 2 − 2 − 2 − 1 1 − 1 2 2 3 (1 + q)t2c qtX X + t( b b q b q2 2b t3 b qt3)X2+ 1 2 1 − 1 − 1 − 1 − 1 − 1 2 ( a q2 + a qt3)X X +(1+ q)t2b tX X + (a q3 + a qt3)X X + − 1 1 2 3 1 3 1 1 1 3 2 t( c q2 c q3 c q4 + c t3 +2c qt3)X2 +(1+ q)t2(1 + t)(1 t + t2)X3+ − 1 − 1 − 1 1 1 3 − 2 (1 + q)t(q3 t3)X X X (1 + q)t(1 + t)(1 t + t2)qX X X + (q3 t3)(1 + q)tX3. − 2 3 1 − − 3 2 1 − 3 4.3. Algebra as singular limit of an Etingof-Ginzburg Calabi-Yau algebra. In thisUZ section we prove some further nice properties of the algebra by showing that it is isomorphic to a singular limit of an Etingof-Ginzburg Calabi-YauUZ algebra. Indeed, the specialisation of relations (4.31) with

2 (d) 2 (d) 2 (d) (q 1)ǫ1 (q 1)ǫ2 (q 1)ǫ3 a1 = − , b1 = − , c1 = − (4.34) √q √q √q a =Ω (1 q), b =Ω (1 q), c =Ω (1 q), t =0, 2 1 − 2 2 − 2 3 − gives the commutation relations (1.5). The following result proves the third statement in Theorem 1.5:

Proposition 4.1. The cubic Casimir Ω4 deﬁned in (1.8) is a special limit of the Etingof-Ginzburg central element ΩEG.

Proof. To deduce the central element Ω4 as a limit of ΩEG, we ﬁrst need to introduce 2 a quadratic term X1 in ΩEG by applying the commutation relations (4.31). Then, 1 2 3 by taking the limit as t 0 of t (ΩEG a1a2(q + t )) we obtain: (4.35) → − Ωt=0 := (q2 1)qX X X (q+1)(a qX +b X +c qX ) a q2X2 b X2 c q2X2. EG − 3 2 1− 2 1 2 2 2 3 − 1 1 − 1 2 − 1 3 t=0 The specialisation of ΩEG with (4.34) is a central element in the algebra that 2 UZ coincides with (q 1)√qΩ4. −

From this perspective, one can specialise the potential ΦEG +ΨEG with the choice of parameters (4.34). In this way, one obtains precisely the potential (1.7). This potential can be decomposed as Φ =ΦSP +Ψ where UZ UZ Φ = X X X qX X X C X ,X ,X SP 1 2 3 − 2 1 3 ∈ h 1 2 3i♮ is a homogeneous degree 3 potential that yields the skew polynomial algebra of three variables X1,X2,X3 (1.6) and 2 (q 1) (d) 2 (d) 2 (d) 2 (d) (d) (d) Ψ = − ǫ1 X1 + ǫ2 X2 + ǫ3 X3 + (q 1)(Ω3 X3 +Ω1 X1 +Ω2 X2), UZ √q − is the specialisation of ΨEG with the choice of parameters (4.34). Therefore we have the following result from which the second statement of Theorem 1.5 follows automatically: 20 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

Proposition 4.2. The associative algebra q A := C X1,X2,X3 / ∂1Φ , ∂2Φ , ∂3Φ . ΦUZ h i h UZ UZ UZ i coincides with and is a non-homogeneous 3-Calabi-Yau Koszul algebra. UZ Proof. To prove that Aq coincides with we simply observe that the cyclic ΦUZ UZ derivatives of the potential Φ give precisely the ﬁrst three expressions in (1.5). q UZ To prove that AΦUZ is a 3-Calabi-Yau Koszul algebra we cannot apply Theorem 3.4.5 of [15] directly to the cubic potential (1.7) because of the fact that limit t 0 is singular. Instead, we use the fact that, as proved in Proposition 3.6, this algebra→ is a PBW deformation of the 3-Calabi-Yau Koszul algebra AΦSP with potential ΦSP and apply Theorem 3.1 in [4] that states that a non-homogeneous graded 3-algebra is a Calabi-Yau Koszul algebra if the homogeneous part is a 3-graded Calabi-Yau Koszul algebra.

Then we can prove the following: q Theorem 4.3. The quotient algebra AΦUZ /(Ω ) is a non-commutative deforma- 1,0 UZ 1,0 tion of the Poisson quotient A 1,0 /(φ ) of the algebra A 1,0 , where φ denotes φǫ,ω ǫ,ω φǫ,ω ǫ,ω τ,t the potential φa,b,c,d with the parameter choice 4.28, and the following commutative diagram holds:

/o/o ﬂ./o/o def./o/o q (4.36) A 1,0 / A = φǫ,ω ΦUZ UZ

1,0 fl./o /o def./o q 0 A 1,0 / φ / A / Ω4 Ω . φǫ,ω h ǫ,ωi ΦUZ h − 4i Proof. The upper horizontal arrow is a direct consequence of Proposition 4.2. The left vertical downward arrow is the Poisson algebra morphism between the Jacobian Nambu–Poisson algebra A 1,0 on C[x , x , x ] and the Poisson algebra of functions φǫ,ω 1 2 3 1,0 C3 on the affine cubic surface φǫ,ω = 0 in . The lower horizontal and right vertical downward arrows are constructed by combining the main ideas of Theorem 3.4.5 of [15], Theorem 3.7, Propositions 3.6 and 4.1 and Theorem 1.5. We remark that (after the computation of the central element by our singular limit (4.1) the flatness argument verification is a routine procedure practically verbatim to the demonstration of propositions 7.3.1, Corollary 7.4.1 and Lemma 7.4.2 in [15].

4.4. Generalised Etingof-Ginzburg cubics. We now replace the homogeneous part Φ given in (4.29) by Φ C X ,X ,X EG α,β,γ ∈ h 1 2 3i♮ 1 (4.37) Φ = X X X qX X X (αX3 + βX3 + γX3) α,β,γ 1 2 3 − 2 1 3 − 3 1 2 3 and consider the family of ﬁltered algebras Aq with generators X ,X ,X Φα,β,γ 1 2 3 subject to the relations X X qX X = γX2, 1 2 − 2 1 3 X X qX X = αX2, 2 3 − 3 2 1 X X qX X = βX2. 3 1 − 1 3 2 QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 21

Due to the results in [22, 23] we know that algebras with homogeneous potentials from (4.37) are non-commutative Koszul 3-Calabi-Yau for certain choices of the parameters α,β,γ but they are not always PBW or PHS.

4.4.1. Digression. Here we list some alternative definitions of the PBW property used in the literature:2 Definition 4.4. The associative filtered algebra A is a PBW-algebra if 1 (1) The algebra A is a Koszul and has Poincaré-Hilbert series PA(t)= (1 t)n . − (2) The elements xi1 xi2 ...,xin , where i ,...,i Z, form a linear basis. 1 2 n 1 n ∈ (3) There is an ordering on generators x1,...,xn w.r.t. which the defining relations form a Gröbner basis. (4) The associated graded algebra is canonically isomorphic to the algebra generated by the homogeneous parts of quadratic relations. For example, the algebra of commutative polynomials satisfies (2) and in the case n = 3 is a PHS algebra. Note that the fourth definition implies that any homogeneous algebra automatically has the PBW property.

Example 4.5. Let A be the quantum algebra given by three generators X1,X2,X3 and three relations 2 2 2 X3 + aX1X2 + bX2X1, X2 + aX3X1 + bX1X3, X1 + aX2X3 + bX3X2 and the parameters (a,b) = (0, 0), (a3,b3) = (1, 1), (a + b)3 = 1. 6 6 6 − This algebra (number P1, table VI in [23]) is PBW with respect to the definitions (1) (2) and (4) in 4.4 but not PBW in sense of the definition (3). Conversely, the algebra B given by three generators Y1, Y2, Y3 and three relations Y Y + bY Y , Y Y + bY Y , Y Y + bY Y , b =0 1 2 2 1 3 1 1 3 2 3 3 2 6 (which is number PII, table VI in [23]) is a PBW-algebra for all definitions in 4.4 4.4.2. Calabi-Yau-Koszulity and PBW- properties of algebras whose potential is non homogeneous. Here we consider algebras whose potential has homogeneous cubic part of Φα,β,γ as well as non-homogeneous terms. Namely, we extend the family of algebras Aq by introducing the potential Φ +Ψ and considering ΦEG+ΨEG α,β,γ EG the family Aq whose relations take the form Φα,β,γ +ΨEG X X qX X γX2 + c X + c =0, 1 2 − 2 1 − 3 1 3 2 (4.38) X X qX X αX2 + a X + a =0, 2 3 − 3 2 − 1 1 1 2 X X qX X βX2 + b X + b =0. 3 1 − 1 3 − 2 1 2 2 Inspired by B. Shoikhet [49], we call this generalised algebra family by Etingof- Ginzburg type algebras. The generalised Etingof-Ginzburg algebra (4.38) is a Koszul, 3-Calabi-Yau for the cases when all constants α,β,γ are equal and non-zero, or only one of them is zero, or if two of the constants are equal and non-zero but q = 1 ([23], Table VIII).

2We are in debt to Natalia Iyudu for her patient explanation and clarification of different definitions of PBW property. 22 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

Below, for γ = 0 we have computed the central element. We stress that the corresponding Etingof-Ginzburg algebras are not Calabi-Yau for generic values of q, α and β. Lemma 4.6. For γ =0, the element (4.39) Ω = q(1 + q)( 1+ q3)X X X + q3(1 + q)αX3 +(1+ q)βX3 a q2(1 + q + q2)X2 GEG − 3 2 1 1 2 − 1 1 + c q(1 + q)αβX X b (1 + q + q2)X2 c q2(1 + q + q2)X2 1 2 1 − 1 2 − 1 3 − qX (a (1+2q +2q2 + q3)+ b c α)+ X ( b (1+2q +2q2 + q3) a c qβ) − 1 2 1 1 2 − 2 − 1 1 − 2 3 2 qX3(c2(1+2q +2q + q )+ c1αβ) is a central element in the algebra (4.38). q As already mentioned, the algebra AΦEG+ΨEG is a non-commutative Calabi-Yau algebra. Moreover, in [15] it is shown that the Hilbert-Poincarépolynomial of the algebra Aq is 1 , i.e. this is a PHS-algebra. Conversely, as follows ΦEG+ΨEG (1 t)3 − q from Example 4.5, the homogeneous degree 3 part A d of this algebra is not a Φt,0,0,0 PBW-algebra in the sense of (4) in Definition 4.4. In the next subsection we discuss a known example of the generalised Etingof– Ginzburg corresponding to α = β = γ = 0 for which the Etingof-Ginzburg type algebra is “good” Koszul Calabi-Yau. 4.5. Odesskii algebra of Sklyanin type. In [34], Odesskii defined a quadratic algebra Oq with three generators X1,X2,X3 satisfying the following relations: (4.40) X X qX X = X ; X X qX X = X ; X X qX X = X , 1 2 − 2 1 3 2 3 − 3 2 1 3 1 − 1 3 2 and proved that, for generic q, the center (Oq) is generated by the following element Ωq := (q2 1)X X X + X2 + q2X2Z+ X2. When q 1 the algebra tends − 1 2 3 1 2 3 → to the universal enveloping U(sl2). Odesskii called the algebra Oq a Sklyanin type algebra.

Theorem 4.7. The Odesskii algebra Oq is a PBW deformation of the 3-Calabi-Yau Koszul algebra of skew polynomials deﬁned by the potential 1 (4.41) Φ := Φ (X2 + X2 + X2) C X ,X ,X . O SP − 2 1 2 3 ∈ h 1 2 3i♮ Proof. This algebra is a PBW-algebra in the sense of all deﬁnitions in 4.4 because of the good PBW-properties in all senses of its homogeneous degree 3 part (see second case in the Example 4.5). To check these properties we again apply the Theorem 3.1 in the case N = 2 of R. Berger et R. Taillefer [4]. Remark 4.8. This algebra is related to the following version of a quantised universal enveloping algebra for sl2 ([29]): make a rotation in the (X1,X2) plane: X X ; X X ; X X 1 →− 2 2 → 1 3 → 3 and then the rescaling 1 1 1 (4.42) X (q q− )X ; X (q q− )X ; X (q q− )X , 1 → − 1 2 → − 2 3 → − 3 maps the Odesskii algebra to the algebra with relations 1 (4.43) qX X X X = (q q− )X ; 1 2 − 2 1 − 3 QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 23

1 qX X X X = (q q− )X ; 2 3 − 3 2 − 1 1 qX X X X = (q q− )X 3 1 − 1 3 − 2 and with the Casimir Ω˜ := qX X X + q2X2 + X2 + X2. O − 1 2 3 1 2 3 Remark 4.9. This quantum Casimir cubic goes to the famous Markov cubic in the limit q 1. → 4.6. Sklyanin algebra with three generators. One of the most famous examples of a 3-Calabi-Yau algebra is the graded associative algebra Q3( ,a,b,c) which is related to a (possibly degenerate or singular) elliptic curve E E (4.44) Q ( ,a,b,c)= C X ,X ,X /J 3 E h 1 2 3i Φ with J = aX X +bX X +cX2, aX X +bX X +cX2, aX X +bX X +cX2 , Φ h 2 3 3 2 1 3 1 1 3 2 1 2 2 1 3 i where (a,b,c) C3 are some parameters. This algebra is a special sub-case of the ∈ b c one generated by ΦEG with q = a and t = a . Artin and Schelter [2] proved that, if the parameters (a,b,c) C3 define the homogeneous coordinates of a point in , this algebra satisfies∈ the Poincare- Birkhoff-Witt condition for all definitionsE in 4.4 except (3) and hence it can be considered as a deformation of the polynomial ring C[x1, x2, x3]. For this reason, this algebra is often called the Artin–Schelter–Tate–Sklyanin algebra with three generators, but in this paper, for brevity, we call “Sklyanin algebra” any graded associative algebra with quadratic relations which satisfies the Poincare-Birkhoff-Witt or PHS-conditions and that can be considered as a deformation of the polynomial ring C[x1, x2, x3]. Iyudu and Shkarin [22] have proved that this algebra is a CY algebra. 4.7. Generalised Sklyanin algebras with three generators. Iyudu and Shkarin ([22]) introduced the generalised Sklyanin algebra with three generators as the following quotient of the free associative algebra (4.45) Q˜ (a,b,c,α,β,γ)= C X ,X ,X /J 3 h 1 2 3i GS where J = X X aX X αX2, X X bX X βX2, X X cX X γX2 , GS h 2 3 − 3 2 − 1 3 1 − 1 3 − 2 1 2 − 2 1 − 3 i where (a,b,c,α,β,γ) C6 is a generic set of complex constants. These generalised∈ Sklyanin algebras are not always potential and in fact, for generic (a,b,c,α,β,γ) they have neither good PBW-properties nor Koszul properties. However, for special values of the parameters they do and a complete classification is given in the following result [22]: Theorem 4.10. The generalised Sklyanin algebra is PHS if and only if at least one of the following conditions is satisfied: (1) For a = b = c =0 and (a3,αβγ) = ( 1, 1) - this case includes the quadratic 6 α 6 − Sklyanin algebra Q3( , a, c, 3 ). (2) For (a,b,c) = (0, 0, 0)E and either α = β = a b =0 or γ = α = c a =0 or β = γ = b6 c =0. − − − 24 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

(3) For a specific choice of all parameters in terms of a root of unity, it is a “finite” algebra which is out of our interest. (4) For a = b = c =0 and αβγ =0. This algebra is potential without the cubic 6 term X1X2X3 and is out of our interest. (5) For α = β = γ = 0 and (a,b,c) = (0, 0, 0), this is the case of the skew polynomial algebra. 6 In all these cases, the generalised Sklyanin algebra is potential and Koszul. The potential can be written as follows: 1 Φ = (αX3 + βX3 + γX3)+ãX X X + ˜bX X X , GS 3 1 2 3 1 2 3 2 1 3 where a˜ and ˜b depend on a,b,c,q. 4.8. Generalised Sklyanin-Painlevépotential. Motivated by the idea of merg- ing together generalised Sklyanin algebra and our algebra (1.5), we consider the following generalisation of the potential of Etingof and Ginzburg to include the first two cases of Theorem 4.10:

(4.46) Φ = ΦGS +ΨEG For the choice of parameters as in the cases of Theorem 4.10, the algebra q := C X ,X ,X /J, A h 1 2 3i where J = ∂ Φ, ∂ Φ, ∂ Φ h X1 X2 X3 i is a generalised Sklyanin-Painlev´ealgebra (1.9) and gives a PHS- or PBW-type 3-Calabi-Yau deformation of C X ,X ,X /J . h 1 2 3i ΦGS For special choices of the parameters α,β,γ,a,b,c,a1,b1,c1,a2,b2,c2, the space of objects q := q/[ q, q] appears in the Physics literature to which section 6 A♮ A A A is dedicated. Remark 4.11. It is interesting to observe the correspondence between the conditions on the constants α, β and γ in the generalised Etingof-Ginzburg algebras and the semi–classical conditions on the existence Poisson-Nambu 3d polynomial algebras in the the paper of L. Vinet and A. Zhedanov ([55]) where they classify various Poisson analogues of the Askey-Wilson algebras AW(3). This correspondence could be behind the fact that the potential and the central element in the generalised Etingof-Ginzburg algebras are, in general, diﬀerent, despite having the same semi-classical limit.

5. Poisson structures and degenerations of elliptic curves Motivated by the observation by M. Gross, P. Hacking and S.Keel (see Example 6.13 of [21]) that the family associated to (2.13) is a log-symplectic Calabi-Yau variety, or in other words, that the projective completion Y of (2.13) with the cubic divisor D given by a triangle of lines, is an example of a Looijenga pair, in this section we study∞ the degenerations of a certain class of Looijenga pairs (Y,D). In this context we need to ﬁx some notation and assumptions to make our discussion clear. We consider the polynomials φ C[x1, x2, x3] of the form (1.2), or (1.3) or belonging to Table 1. We list all such∈ polynomials in the ﬁrst column of Table 3. QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 25

δ Polynomials φ φ weights ∞ x6 x3 x2 1 x6 x3 x2 1 + 2 + 3 + τx x x + η x5 + + ω, 1 + 2 + 3 + τx x x 6 3 2 1 2 3 5 1 ··· (1, 2, 3) 6 3 2 1 2 3 4 4 2 2 4 4 2 x1 + x2 + x3 + τx x x + η x3 + + ω, x1 + x2 + x3 + τx x x 4 4 2 1 2 3 3 1 ··· (1, 1, 2) 4 4 2 1 2 3 x3 x3 x3 3 x3 x3 x3 1 + 2 + 3 + τx x x + η x2 + + ω, 1 + 2 + 3 + τx x x 3 3 3 1 2 3 2 1 ··· (1, 1, 1) 3 3 3 1 2 3 1 x x x + x5 + x2 + x2 + η x4 + + ω, x x x + x5 + x2 1 2 3 1 2 3 4 1 ··· (2, 5, 3) 1 2 3 1 2 2 x x x + x4 + x2 + x2 + η x3 + + ω, x x x + x4 + x2 1 2 3 1 2 3 3 1 ··· (1, 2, 1) 1 2 3 1 2 3 x x x + x3 + x3 + x2 + η x2 + + ω, x x x + x3 + x3 1 2 3 1 2 3 2 1 ··· (1, 1, 1) 1 2 3 1 2 3 x x x + 3 (ω x ǫ x2 )+ ω x x x 1 2 3 k=1 k k − k k 4 (1, 1, 1) 1 2 3 P Table 3. del Pezzo surfaces as Loojenga pairs - in the last row we dropped the index (d).

In each case, the projective completion φ of φ in the weighted projective 3 M M spaces WP are del Pezzo surface of degree δ [16, 15]. We denote by (x0,...,x3) the weighted homogeneous coordinates in WP3. We list the degree δ and the weights of the variables (x1, x2, x3) in the second column - we always assume the weight of the homogeneous coordinate x0 to be 1. For each polynomial φ C[x1, x2, x3] in Table 3, we take the weighted homogeneous part φ and list it∈ in the third column. The equation φ = 0 defines a ∞ 2 ∞ projective curve in WP . The pair ( φ,D ) is a Looijenga pair and φ D is the affine surface C3. M ∞ M \ ∞ Mφ ∈ The projectivisation P φ of φ is a projective manifold of dimension 1 embedded in P2 by the linear systemM givenM by sections of a line bundle of degree δ - for degenerated cubic divisors of del Pezzo degree 2 and 1 such sections are expressed via Gross-Hacking-Keel θ-functions. The coordinate ring of φ D is C[x1, x2, x3]/ φ , which corresponds to the cone over the projectivisationM P\ ∞ P2, namely h i Mφ ∈ 0 k C[x , x , x ]/ φ = H (P ,L⊗ ), 1 2 3 h i ⊕k Mφ where L is the trivial bundle of degree δ. By taking the generalisation of the Poincaréresidue for weighted projective spaces (see for example, [13]) of the global 3-form in WP3 along the divisor D , one obtains a symplectic form on the quo- ∞ tient C[x1, x2, x3]/ φ which descends from the Nambu bracket restricted to the symplectic leaves φh=i 0. In this Section we carry out the above construction for each φ in Table 3. We also consider special cases and degenerations, namely singular limits obtained by rescaling the weighted homogeneous coordinates and taking limits of such rescaling to infinity. We show that such degenerations correspond to rational degenerations of elliptic curves. 26 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

5.1. Degenerations of the Sklyanin algebra with three generators. In this 3 3 3 x1 x2 x3 subsection we consider φ = 3 + 3 + 3 +τx1x2x3, a special case of the third row of Table 3. This case is∞ related to the semi–classical limit of the Sklyanin algebra (4.44); namely, take a,b such that a + b is proportional to 1 q, the semi—classical c − limit q3( , τ) (where τ = 3 ) of the Sklyanin algebra Q3( ,a,b,c) carries a Poisson structureE (which is also called Poisson Sklyanin algebra).E In [35] and [40] it was shown that this Poisson algebra belongs to a family of Poisson structures on the moduli space of parabolic vector bundles of degree 3 and rank 2 on the projective 2 space P . The explicit expression for the elliptic Poisson brackets of q3( , τ) is the natural one carried by the family of the Hesse cubics E 1 (5.47) φ = (x3 + x3 + x3)+ τx x x =0 τ 3 1 2 3 1 2 3 that define the embedding of in P2. Namely the quadratic brackets on the affine space C3 which define a quadraticE Poisson algebra structure on C 0 k Aφτ = [x1, x2, x3]/φτ = k 0H (φτ ,L⊗ ) ⊕ ≥ and L is the degree 3 line bundle over the cubic curve φτ are: (5.48) x , x = x2 + τx x ; x , x = x2 + τx x ; x , x = x2 + τx x . { 1 2} 3 1 2 { 2 3} 1 2 3 { 3 1} 2 3 1 It a straightforward computation to check that the algebra q ( , τ) is invariant 3 E under the Heisenberg group H3 and unimodular (see [38]). 5.1.1. Rational degenerations of Sklyanin Poisson algebra and triangular divisor of Painlevéprojective surfaces. A. Odesskii in [33] proposed a description of all rational degenerations for a generalisation of elliptic algebras known as Sklyanin– Odesskii–Feigin algebras, and their semi–classical counterparts - namely rational Poisson quadratic algebras. We shall restrict ourselves to one example of it in the case of the Poisson elliptic algebra q3( , τ). It is shown in [33] that the center of the rational degeneration R1( 2 ) of theE Sklyanin algebra Q ( ,a,b,c) is generated 3 − 3 3 E by one polynomial of degree 3 in P2 1 φ˜ = y3 + y y y . 3 2 1 2 3 Indeed, if we take the Casimir element φ of q3( , τ) given by the Hesse cubic (5.47) and take the rational limit τ which givesE us the triangle configuration in Figure 1 → ∞ x =0 x =0 x =0 { 1 }∪{ 2 }∪{ 3 } in the coordinates yi,i =1, 2, 3 defined as:

y1 = √τx1, y2 = x2, y3 = √τx3, we obtain φ˜. The same triangle conﬁguration is the divisor at inﬁnity of the projective com- .(pletion φ¯ ֒ P3 of the general Painleve cubic (2.13 P → Remark 5.1. It is clear that, in the limit τ , the Poisson brackets (5.48) give the cluster Poisson structure ([18]) → ∞ x , x = x x ; x , x = x x ; x , x = x x , { 1 2} 1 2 { 2 3} 2 3 { 3 1} 3 1 but in the degenerated coordinates y1,y2,y3 these brackets read y ,y = y y ; y ,y = y y ; y ,y = y2 + y y . { 1 2} 1 2 { 2 3} 2 3 { 3 1} 2 3 1 QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 27

Because these are brackets on C3, they define a quadratic Poisson algebra structure on 0 k A ˜ = C[y1,y2,y3]/φ˜ = k 0H (φ,L˜ ⊗ ) φ ⊕ ≥ 1 3 where L is the degree 3 line bundle over the cubic divisor 3 y2 + y1y2y3 = 0 which is 1 2 the union of the line y2 = 0 and the conic 3 y2 +y1y3 = 0. The rational degeneration deforms the cluster Poisson structure. We will consider the quantum version of this in subsection 6.4. 5.1.2. Elliptic curves in weighted projective spaces, related Sklyanin Pois- son structures and their rational degenerations. We deal first with the polynomial φ in the third row of Table 3. As discussed in subsection 4.2, it is convenient to write this polynomial in the form t 1 φτ,t = τx x x + (x3 +x3 +x3)+ (a x2 +b x2 +c x2)+a x +b x +c x +d. a,b,c,d 1 2 3 3 1 2 3 2 1 1 1 2 1 3 2 1 2 2 2 3 2 The projectivisation PM τ,t of the hypersurface M τ,t is a curve in P and φa,b,c,d φa,b,c,d M τ,t can be seen as a line-bundle over PM τ,t . When t = 1, a = b = c = φa,b,c,d φa,b,c,d d = 0 the surface M τ,1 is an affine cone over a normally embedded elliptic curve φ0,0,0,0 in P2 of degree 3 given by the homogeneous cubic

1 3 3 3 P2 φ = τx1x2x3 + (x1 + x2 + x3)=0 . ∞ 3 ⊂

This cone surface M τ,1 is an example of a simple elliptic Gorenstein singu- φ0,0,0,0 larity (E6 case corresponding to the elliptic singularities list). Note that the same formula for φ also defines a hypersurface in C3 with a triple point singularity in 0. Let use now deal with the first two lines of Table 3. Denote by φ1,1,2 and φ2,1,3 the φ in the second and first row respectively. The∞ surface 1 1 1 (5.49) φ = τ x x x + x4 + x4 + x2 =0 C3 1,1,2 1 1 2 3 4 1 4 2 2 3 ⊂ 3 has a double point in C that is an elliptic Gorenstein singularity of type E7. It defines the affine cone over a homogeneous degree 4 elliptic curve in weighted projective space WP1,1,2 defined by the same equation φ1,1,2 = 0. Similarly, thee surface of type E8

1 3 1 6 1 2 C3 (5.50) eφ2,1,3 = τ2x1x2x3 + x1 + x2 + x3 =0 3 6 2 ⊂ which is the aﬃne cone over a homogeneous degree 6 elliptic curve in weighted projective space WP2,1,3 deﬁned by the same equation φ2,1,3 = 0. From an algebraic point of view the coordinate rings Aφ discussed in subsection

4.2, both Aφ1,1,2 and Aφ2,1,3 are graded rings such that 0 k (1) Aφ = C[x1, x2, x3]/φ = k 0H (φ, L⊗ ) where L is the degree 3 line bundle ⊕ ≥ over the cubic curve φ and the sections of L form the linear system3 deﬁning ;the embedding φ ֒ P2 →

3An explicit construction of linear systems deﬁned by sections of L for degree 2 and 1 in terms of appropriate theta functions similar to this case can be found, for example, in the paper [46]. 28 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

0 k C (2) Aφ , , = k 0H (φ1,1,2,L⊗ )= [x1, x2, x3]/φ1,1,2, where L is the degree 1 1 2 ⊕ ≥ 2 line bundle over the nodal curve φ1,1,2 and the sections of L define the .embedding φ1,1,2 ֒ WP1,1,2 0→ k C (3) Aφ , , = k 0H (φ2,1,3,L⊗ )= [x1, x2, x3]/φ2,1,3 where L is the degree 2 1 3 ⊕ ≥ 1 line bundle over the nodal curve φ2,1,3 and the sections of L define the . embedding φ ֒ WP 2,1,3 → 2,1,3 In the seminal paper by Gross, Hacking and Keel [21], the notion of vertex of degree n, Vn, was introduced. These surfaces Vn are the most important singularities for moduli of smooth surfaces. We shall be specially interested in the degenerate cases of vertices V1,2 of degree 1 and 2 which can be defined as follows: V = Spec(C[y ,y ,y ]/(y y y y2 y2)) 1 1 2 3 1 2 3 − 1 − 3 and 2 2 2 V2 = A A = Spec(C[y1,y2,y3]/(y1y2y3 y )). y1,y2 ∪ y2,y1 − 3 The first surface V1 is the affine cone over a nodal cubic curve in the weighted projective space WP2,1,3 and the second, V2, is the affine cone over the union of two rational curves embedded in the weighted projective space WP1,1,2. 3 Both affine cones V1 and V2, being affine surfaces in C carry the natural Pois- 2 2 son Jacobi unimodular structures defined by the potentials φ1 = y1y2y3 y1 y3 2 − − and φ2 = y1y2y3 y3 respectively. P.Bousseau has identified these Jacobian Pois- son brackets with− the natural Poisson structures on classical Gross-Siebert theta- functions (Th. 21 and Corollary 22 of [6]), defined in terms of so called “broken lines” data.

Proposition 5.2. The vertex surfaces V1 and V2 coincide with degenerations of the aﬃne cones φ2,1,3 and φ1,1,2 respectively. Their Jacobian Poisson algebras are isomorphic toM degenerationsM of elliptic Poisson Sklyanin algebras deﬁned by the weighted cubic Casimirs φ2,1,3 and φ1,1,2 respectively. Proof. We apply the same procedure of degeneration as above, namely we rescale

1/3 y2 1/2 x1 y13 , x2 = 1/2 1/3 , x3 = y32 , → −τ22 3 and take the limit τ to obtain 2 → ∞ φ = y3 + y2 y y y . 2,1,30 1 3 − 1 2 3 The corresponding Jacobian Poisson brackets read as (5.51) y ,y =2y y y , y ,y =3y2 y y , y ,y = y y { 1 2} 3 − 1 2 { 2 3} 1 − 3 2 { 3 1} − 1 3 and deﬁne a Posson algebra structure on the ring C 0 k (5.52) Aφ , , := [y1,y2,y3]/φ2,1,3 = k 0H (φ2,1,3 ,L⊗ ), 2 1 30 0 ⊕ ≥ 0 ˜ where L is degree 1 line bundle over the singular curve φ2,1,30 = 0, i.e. the rational nodal cubic of arithmetic genus 1 embedded in WP2,1,3 . Then

φ := Spec(Aφ ) M 2,1,30 2,1,30 C3 V is the affine cone in over the singular curve φ2,1,30 = 0, namely 1. Similarly by 1 1 √ x1 = 1/4 y1, x2 = 1/4 y2, x3 = 2y3, −2 √τ1 2 √τ1 QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 29 in the limit τ one has 1 → ∞ (5.53) φ = y2 y y y . 1,1,20 3 − 1 2 3 The corresponding Jacobian Poisson brackets read as (5.54) y ,y =2y y y , y ,y = y y , y ,y = y y , { 1 2} 3 − 1 2 { 2 3} − 3 2 { 3 1} − 1 3 and define a Posson algebra structure on the ring C 0 k (5.55) Aφ , , := [y1,y2,y3]/φ1,1,2 = k 0H (φ1,1,2,L⊗ ), 1 1 20 0 ⊕ ≥ where L is degree 2 line bundle over φ1,1,2 = 0, the union of two rational curves y = 0 and y y y = 0 embedded in WP and 3 3− 1 2 1,1,2 φ = Spec(Aφ ) M 1,1,20 1,1,20 3 is the affine cone in C which is nothing but the 2-vertex V2.

In section 6.3 we provide a quantisation of these two degenerate cases and cal- culate the central elements - the quantisation of the full non-degenerate case can be found in [15]. Remark 5.3. In the next section, we will discuss the appearance of the same degenerated Jacobian Poisson structures as a result of semi–classical limit of the quantum generalized Sklyanin-Painlevé“marginal” deformation of a quiver theory prepotential. Note that in the weighted projective space, there are many different homogeneous polynomials φ of degree 4 that define the same quotient by the Jacobian ideal. For example 1 (5.56) φ˜ = τ x˜ x˜ x˜ + (˜x2 +˜x x˜2 +˜x x˜3)=0 WP , 1,1,2 1 1 2 3 3 3 3 1 1 2 ⊂ 1,1,2 defines the same algebra as φ1,1,2 and 1 (5.57) φ˜ = τ x˜ x˜ x˜ + (˜x3 +˜x3x˜ +˜x2)=0 WP , 2,1,3 2 1 2 3 3 1 2 3 3 ⊂ 2,1,3 defines the same algebra as φ2,1,3. In [36], A. Odesskii and the third author described two non-rational Poisson morphisms between the Poisson algebra of Jacobian type associated with the Hesse cubic (5.48) in the variables and the two homogeneous polynomials φ1,1,2:

3 3 1 1 1 2 − 4 4 − 2 2 (5.58)x ˜1 = x1 x3 , x˜2 = x2x3 x1 , y3 = x3 and φ2,1,3:

1 3 − 2 2 (5.59)x ˜1 = x1, x˜2 = x2x3 , x˜3 = x3 . As discussed in [36], the non-rational Poisson morphisms (5.58), (5.59) have their origin in the Calabi-Yau mirror symmetry dualities ([19]) and the question of their “quantum” interpretation was posed. Because by rescalingx ˜1, x˜2, x˜3 in exactly the same way as x1, x2, x3 one can produce the same rational limits φ2,1,30 , φ1,1,20 , the quantisation produced in subsection 6.3 gives a partial answer to this question by providing a quantisation for some rational limits of φ˜1,1,2 and φ˜2,1,3. 30 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

5.1.3. Degnerate Sklyanin algebras with three generators. The Sklyanin algebra Q ( ,a,b,c) has the following degeneration locus 3 E = (1, 0, 0); (0, 0, 1); (0, 0, 1) (a,b,c) a3 = b3 = c3 . D { }⊔{ | }

Following [50], we call degenerate Sklyanin algebra the algebra Q3( ,a,b,c) with (a,b,c) D. E It was∈ proven by P. Smith that such a degenerate Sklyanin algebra is isomorphic to C u,v,w /J where the ideal J is J = u2 = v2 = w2 = 0 if a = b, and J = uv =h vw =iwu =0 if a = b. In the semiclassicalh limit the latteri case corresponds toh φ = uvw, which isi the6 decorated character variety of π (P1 z ,z ,z ) [10]. 1 \{ 1 2 3} Remark 5.4. The latter model has a quiver representation with potential Q = uu∗ + vv∗ + ww∗ uvw wvu [8]. − −

6. Non-commutative cubics and QFT deformations There is an interesting similarity between the formulae for the quantum potential Φ deﬁned in (4.46) and the non-commutative potentials describing the marginal and relevant deformations of the N = 4 super Yang-Mills (SYM) theory in four dimensions with gauge group U(n) (see [3] for a physical background) This theory is written in terms of the N = 1 SYM theory with three adjoint chiral super-ﬁelds X1,X2,X3 coupled by the potential:

Φsmooth = gTr([X1,X2]X3) with coupling constant g, where, following the physics literature Tr denotes the map A A . From now on we drop Tr, i.e. we denote potentials and their images → ♮ in A♮ with the same symbol. The moduli space of supersymmetric gauge theories is an important and rather well-studied object (a mathematical account of this theory can be found in the recent paper of C. Walton [56]). The marginal deformations, which preserve some conformal symmetry, of the N = 4 Superconformal Field Theory have many interesting applications. In particular, within the framework of the AdS/CFT correspondence, they have a nice Supergravity dual descriptions. If one chooses to preserve N = 1 Super Conformal Field Theory then the moduli space of the marginal deformations is given by the potential: 1 (6.60) Φ = X X X qX X X + λ(X3 + X3 + X3). marg 1 2 3 − 2 1 3 3 1 2 3 Another important class of deformations is provided by relevant deformations which describes the theory away from the Ultra-Violet conformal ﬁxed point:

2 2 2 Φrel = m1X1 + m2(X2 + X3 )+ dkXk. Xk The structure of the vacua of D-brane gauge theories relates to Non-Commutative Geometry via the potentials Φ by so called F term constraints: phys − ∂Φ (6.61) phys =0, k =1, 2, 3 ∂Xk QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 31 where Φphys = Φmarg +Φrel. This gives rise to the following non homogeneous relations: X X qX X = ΛX2 m X d 1 2 − 2 1 − 3 − 2 3 − 3 (6.62) X X qX X = ΛX2 m X d  2 3 3 2 1 1 1 1 X X − qX X = −ΛX2 − m X − d  3 1 − 1 3 − 2 − 2 2 − 2 This algebra is a particular case of the algebra Aq studied in subsection  ΦEG+ΨEG 4.4 for α = β = γ = Λ, a1 = m1, b1 = c1 = m2, a2 = d1, b2 = d2, c2 = d3, or in other words, of the general− algebra q introduced in subsection 4.8. A Remark 6.1. We precise how this deformation algebra relates to previously studied: 1 If Λ = 0 and m1 = m2 = 2 , ei = 0, i = 1, 2, 3 then we have the • potential of (4.41) and this algebra− coincides with the Odesskii degeneration of Sklyanin algebra in subsection 4.5; 1 (V I) We see that for Λ = 0, m1 = m2 = √q(q− q) and di = (1 q)Ωi the • Poisson algebra (6.62) has the form (1.5). − − If Λ= c and q = b/a, then this Poisson algebra coincides with a deforma- • a tion of the quadratic Sklyanin Poisson algebra with three generators q3( ). The latter can be obtained from this deformation by setting the massessE to zero: m1 = m2 =0. 6.1. Semi-classical limits. We now take the semi-classical limit of (6.62) and compare it with the cubic surfaces φ := Spec(C[x1, x2, x3]/ φ = 0 ), for φ in table 1. These cubics are endowed withM the natural Poisson brackeh t (2.11).i By the correspondence principle

[X1,X2] lim = x1, x2 , q 1 1 q { } → − and, applying the algebra relations [X ,X ] = (q 1)X X ΛX2 m X d 1 2 − 2 1 − 3 − 2 3 − 3 so that Λ 2 m2X3 d3 x1, x2 = x1x2 lim X3 lim + lim , { } − q 1 1 q − q 1 1 q q 1 1 q → − → − → − and similarly

Λ 2 m1X1 d1 x2, x3 = x2x3 lim X1 lim + lim , { } − q 1 1 q − q 1 1 q q 1 1 q → − → − → − Λ 2 m2X2 d2 x3, x1 = x1x3 lim X2 lim + lim . { } − q 1 1 q − q 1 1 q q 1 1 q → − → − → − By a slight abuse of notation, we denote the classical masses again by m1,m2, di the classical limit of Λ by λ and put δi = limq 1 , so that the Casimir function → 1 q for this Poisson algebra is − λ φ (x , x , x )= x x x m x2 m (x2+x2) (x3+x3+x3)+δ x +δ x +δ x . cl,tot 1 2 3 1 2 3− 1 1− 2 2 3 − 3 1 2 3 1 1 2 2 3 3 We see that the corresponding Poisson algebras include all interesting families of quadratic-linear-constant Poisson brackets in C[x1, x2, x3] and, in particular, for λ = 0, the family coincides with the Poisson structure on the Painlev´emonodromy 32 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV data cubics. At the same time, by neglecting the terms of degree < 3 in φcl,tot we obtain λ (6.63) φ (x , x , x )= x x x m x2 m (x2 + x2) (x3 + x3 + x3), cl,marg 1 2 3 1 2 3 − 1 1 − 2 2 3 − 3 1 2 3 that is a perturbation of the classical Sklyanin algebra q ( ) (see section 5.1). 3,1 E

6.2. Degeneration of quadratically perturbed q3-Sklyanin brackets and vertex V1,2 varieties. Consider the special case of (6.63) with m2 = 0 and λ = 3 3 : m1

2 1 3 3 3 (6.64) φcl,1(x1, x2, x3)= x1x2x3 m1x1 3 (x1 + x2 + x3). − − m1 This is an example of a central element for the classical Sklyanin algebra perturbed 2 by the quadratic term m1x1. The following proposition is a rephrasing of Proposi- tion 5.2 with the “infinite mass limit” interpretation. Proposition 6.2. There are two “mass re-scalings” of (6.64) such that in the infinite mass limit m1 , φcl,1 degenerates to the potential φ2,1,30 of the vertex V → ∞ 1 Poisson algebra in the first rescaling, and to the potential φ1,1,20 of the Poisson algebra of the vertex V2 in the second rescaling.

Proof. We introduce the coordinates yi,i = 1, 2, 3 connected to x1, x2, x3 by the following relations: y1 y2 x1 = , x2 = , x3 = m1y3, √m1 √m1 so that the Casimir now reads as (y3 + y3) (6.65) φ (y ,y ,y )= y y y y2 y3 + 1 2 . cl,2 1 2 3 1 2 3 1 3 9 − − m1

In the inﬁnite mass limit m1 , (6.65) goes evidentlyp to → ∞ (6.66) φ (y ,y ,y )= y y y y2 y3 cl,3 1 2 3 1 2 3 − 1 − 3 ˜ Note that up to permutations of y1,y2,y3, φcl,3 is the same as φ2130 , and therefore, V C3 as discussed at the end of subsection 5.1.2, the cubic surface 1 = φcl,3 2 3 M ⊂ given by φcl,3(y1,y2,y3)= y1y2y3 y1 y3 can be considered as an aﬃne cone over a singular genus one rational curve− − WP(3, 1, 2). Its coordinate ring Esing ⊂ C[ ]= C[y ,y ,y ]/(y y y y2 y3) Mφcl,3 1 2 3 1 2 3 − 1 − 3 0 is isomorphic to the ring of sections k 0H ( sing , (k)) of a degree 1 line bundle (1)) on the nodal rational curve ⊕ ≥of arithmeticE O genus 1 (see [21] ch.5). This O Esing cone is parametrized by toric theta-functions ϑi, i =1, 2, 3 satisfying the relation 2 3 ϑ1ϑ2ϑ3 = ϑ1 + ϑ3 (see Theorem 2.34 of [21]). Now we come back to the Poisson algebra corresponding to (6.64): 2 2 2 3x3 3x1 3x2 x1, x2 = 3 +x1x2; x2, x3 = 3 2m1x1+x2x3; x3, x1 = 3 +x3x1 { } − m1 { } − m1 − { } m1 which will be written in the degenerated coordinates yi,i =1, 2, 3 as QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 33

(6.67) 3y2 3y2 y ,y = 3y2 +y y ; y ,y = 1 2y +y y ; y ,y = 2 +y y . 1 2 3 1 2 2 3 9 1 2 3 3 1 9 3 1 { } − { } − m1 − { } m1 From this, in the infinite mass limit wep obtain once again (compare withp (5.51)) a perturbed cluster Poisson structure: (6.68) y ,y = 3y2 + y y ; y ,y = 2y + y y ; y ,y = y y { 1 2} − 3 1 2 { 2 3} − 1 2 3 { 3 1} 3 1 which defines the Poisson algebra structure on the coordinate ring of the affine cone over the curve . Esing If, instead, we introduce the coordinatesy ˜i,i =1, 2, 3 connected to x1, x2, x3 by the following relations:

y˜1 x1 = , x2 =y ˜2, x3 = √m1y˜3, √m1 the Casimir now reads as 3 2 1 y˜1 3 3/2 3 (6.69) φcl,4(˜y1, y˜2, y˜3)=˜y1y˜2y˜3 y˜ ( +˜y + m y˜ ). 1 3 3 2 1 3 − − m1 m1

In the infinite mass limit m1 , (6.69) goesp evidently to → ∞ (6.70) φ (˜y , y˜ , y˜ )=˜y y˜ y˜ y˜2 cl,5 1 2 3 1 2 3 − 1 Note that (up to the change of variabley ˜1 = y1,y ˜2 = y3,y ˜3 = y2) φcl,5 is the same as φ˜ , and therefore, as before, the cubic surface V = C3 is an affine 1120 2 Mφcl,4 ⊂ cone over the singular curve ˜ WP(2, 1, 1). Its coordinate ring Esing ⊂ C[ ]= C[˜y , y˜ , y˜ ]/(˜y y˜ y˜ y˜2) Mφcl,4 1 2 3 1 2 3 − 1 0 is isomorphic to the ring of sections k 0H ( ˜sing , (k)) of a degree 2 line bundle ⊕ ≥ E O (1)) on the degenerated curve ˜ which is a union of conic and a line. This O Esing cone is parametrised by Gross-Siebert toric theta-functions ϑi, i =1, 2, 3 satisfying the relation (ϑ ϑ ϑ )ϑ =0 1 2 − 3 3 (see Proposition 40 of [6]). By writing the Poisson algebra corresponding to (6.64) in the new coordinates y˜1, y˜2, y˜3 and taking the infinite mass limit we obtain once again (compare with 5.51) a perturbed cluster Poisson structure: (6.71) y˜ , y˜ =y ˜ y˜ ; y˜ , y˜ = 2˜y +˜y y˜ ; y˜ , y˜ =y ˜ y˜ { 1 2} 1 2 { 2 3} − 1 2 3 { 3 1} 3 1 which defines the Poisson algebra structure on the coordinate ring of the affine cone over the curve ˜ . Esing 6.3. Quantisation of Gross-Siebert theta functions. In [6], P. Bousseau proposed a deformation quantisation for some Poisson algebra H0( , ) structures connected with mirror duals of Looijenga pairs (Y,D) where Y isX aO smoothX projective surface and D some singular anticanonical divisor. Here S is a (possibly singular) family of affine Poisson varieties. As examples he computedX → the deformation quantisation of function algebras on affine r vertex varieties Vr where r is the number of irreducible components of the cubic− divisor D of the A-side of mirror 34 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV correspondence. As we have seen, when r = 1 the 1-vertex variety V1 is exactly the affine cone of the nodal curve embedded in the weighted projective space WP 2,1,3 : V 1 = φ˜ = SpecAφ˜ M 2,1,30 2,1,30 where A is given in (5.52). The Poisson algebra on V is given by the brackets φ˜2,1,3 1 (5.51). V(q) The Proposition 41 in [6] states that the quantum algebra 1 is generated by three generators Yˆ1, Yˆ2 and Yˆ3 with the relations √qˆYˆ Yˆ 1 Yˆ Yˆ =0 3 1 − √qˆ 1 3 √qˆYˆ Yˆ 1 Yˆ Yˆ = (ˆq qˆ 1)Yˆ  2 3 √qˆ 3 2 − 1 −1 3/2 − 3/2 2  √qˆYˆ1Yˆ2 Yˆ2Yˆ1 = (ˆq qˆ− )Yˆ − √qˆ − 3 and that the 0–level of the central element: Ωˆ (Yˆ )= Yˆ Yˆ Yˆ qˆ1/2Yˆ 2 qˆYˆ 3 2,1,3 2 3 1 − 1 − 3 gives the quantisation of the algebra (6.68) (equivalent to (5.51)). As another example, P. Bousseau considered also a deformation quantisation of the function algebra on the 2-vertex V2 related to the mirror dual of the Looijenga pair (Y,D) where the divisor has two connected components, namely for V2 = φ0 , where φ1,1,2 is given in (5.53) and the Poisson algebra is the Jacobian M 1,1,2 0 algebra on C[y1,y2,y3] with the brackets (5.54). The Proposition 40 in [6] describes V(q) ˆ ˆ ˆ the quantum algebra 2 in a similar fashion with three generators Y1, Y2 and Y3 with the relations √qˆYˆ Yˆ 1 Yˆ Yˆ =0 3 1 − √qˆ 1 3 √qˆYˆ Yˆ 1 Yˆ Yˆ =0  2 3 √qˆ 3 2 1 − 1/2 1/2 2  √qˆYˆ1Yˆ2 Yˆ2Yˆ1 = (ˆq qˆ− )Yˆ − √qˆ − 3 and the 0–level of the central element: Ωˆ (Yˆ )= Yˆ Yˆ Yˆ qˆ1/2Yˆ 2 1,1,2 1 2 3 − 3 gives the quantisation of (5.54). The deformational quantisation of the affine Poisson family S elaborated by P. Bousseau was based on Gross–Haking–Keel approach to torXic → mirror conjecture, tropicalization and quantum scattering diagrams and broken lines. On the other hand, if Y in the Looijenga pair (Y,D) is a del Pezzo surface of degree 1, 2 or 3 and the divisor D a nodal cubic, Etingof, Oblomkov and Rains constructed the corresponding quantum algebras as spherical sub–algebras of their generalised DAHA ([16]). In Subsection 4.3, we obtained the generalised Sklyanin–Painlevé algebra as a degeneration of the E6 spherical sub–algebra and fit it within the Etingof–Ginzburg quantisation scheme (see Theorem 4.3). A natural question posed by Bousseaue is to compare his deformation quantisations with the scheme of quantisation proposed by Etingof and Ginzburg. In the next theorem we answer this question in the case of degenerated affine del Pezzo surfaces (the case V1 and V2). Theorem 6.3. The deformation quantisations of the affine Poisson structures on V1,2 obtained in [6] coincide (after a proper rescaling) with the appropriate degenerations of the quantum Sklyanin-Painlevéalgebras defined by relations (1.9). QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 35

Proof. We start by observing that the quantum algebra corresponding to (6.70) is a C degenerate case of the Calabi-Yau algebra X1,X2,X3 /JΦphys with the potential (6.61). Indeed, by analogy with the classicalh case, wei introduce the coordinates Yi,i =1, 2, 3 connected to X1,X2,X3 by the following relations:

Y1 Y2 X1 = , X2 = , X3 = m1Y3, √m1 √m1 to obtain Λ Y 3 + Y 3 Φ = Y Y Y qY Y Y + m3Y 3 + 1 2 + phys 1 2 3 − 2 1 3 3 1 3 3 (6.72) m1 ! 1 2 m2 2 2 p + Y1 + Y2 + m1m2Y3 + e1Y1 + e2Y3 + e3Y2, 2 m1 which is by our discussion a PBW non-homogeneous deformation of the Koszul 3 generalised Sklyanin algebra. By putting Λ = m1− , m2 = 0 and e1 = e2 = e3 = 0, we obtain 1 Y 3 + Y 3 1 Φ = Y Y Y qY Y Y + Y 3 + 1 2 + Y 2, m1 1 2 3 2 1 3 3 3 3 1 − 3 m1 m1 ! 2 and in the limit m1 we obtain p → ∞ 1 3 1 2 Φ (Y )= Y1Y2Y3 qY2Y1Y3 + Y3 + Y1 ∞ − 3 2 and the corresponding quantum algebra C Y , Y , Y /J has relations h 1 2 3i Φ∞ Y3Y1 qY1Y3 = 0 (6.73) Y Y − qY Y = Y  2 3 3 2 1 Y Y − qY Y = Y 2  1 2 − 2 1 3 This algebra has central element  q q2 (6.74) Ωm1 (Y )= Y Y Y + Y 2 + Y 3 0 3 2 1 q2 1 1 q3 1 3 − − and quantises the coordinate ring of the cone over the nodal rational genus 1 curve or the coordinate ring of the aﬃne surface (6.70). But these are the same as (6.73) and (6.74) by setting 3 2 2 1 (1=q ˆ)(ˆq 1) (1+q ˆ+ˆq ) 7 ˆ ˆ 4 ˆ 2 3 5 q = , Y1 = − 1 Y1, Y2 =q ˆ Y2, Y3 = (1 qˆ qˆ +ˆq )Y3. qˆ qˆ4 − −

We can degenerate the algebra (6.73) further by rescaling the variables Y1, Y2, Y3 and taking diﬀerent limits. Namely, setting Y ǫ Y , Y ǫ Y , Y ǫ Y , 1 → 1 1 2 → 2 2 3 → 3 3 we obtain Y Y qY Y = 0 3 1 − 1 3 Y Y qY Y = ǫ1 Y (6.75) 2 3 3 2 ǫ2ǫ3 1  − 2 ǫ3 2  Y1Y2 qY2Y1 = Y3 − ǫ1ǫ2 with central element  q ǫ q2 (6.76) Ωm1 (Y )= Y Y Y + 1 Y 2 + ǫ2ǫ ǫ Y 3 0 3 2 1 q2 1 ǫ ǫ 1 q3 1 3 1 2 3 − 2 3 − 36 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

Imposing ǫ = 1, ǫ = ǫ2, ǫ , in the limit ǫ 0, we obtain 2 1 3 3 3 → Y Y qY Y = 0 1 3 − 3 1 (6.77) Y Y qY Y = 0  2 3 3 2 Y Y − qY Y = Y 2  2 1 − 1 2 3 and the central element is given by  1 2 (6.78) Ω∞(Y )= Y Y Y + Y . 2 3 1 q2 1 3 −

Observe that by choosing different values and limits of ǫ1,ǫ2,ǫ3 in (6.75), we can recognise the algebras given by the super-pontentials of non-commutative Painlevé cubics (PIV and PII) to which the next two subsections are dedicated. 6.3.1. One non-zero mass and PainlevéIV. We consider the deformation provided by addition a single mass term to Φsmooth. The corresponding potential (4.2 of [3]) reads (up to symmetric group Σ3-action): m (6.79) Φ = X X X qX X X X2. 1m 1 2 3 − 2 1 3 − 2 1 The corresponding ideal is defined by (6.80) X X qX X = 0; X X qX X = mX ; X X qX X =0 1 2 − 2 1 2 3 − 3 2 1 3 1 − 1 3 Taking the Poisson limit q 1 one gets the cubic Casimir : → m φ (x , x , x )= x x x x2. cl,PIV 1 2 3 1 2 3 − 2 1 Once again, to link with some of our Painlevécubics ( in the single mass case it will be the PIV cubic) we need to add the linear terms: m (6.81) Φ = X X X qX X X X2 + d X + d X + d X . 1,m 1 2 3 − 2 1 3 − 2 1 1 1 2 2 2 3 Taking d2 = d3 one gets (6.82) X X qX X = d ; X X qX X = mX +d ; X X qX X = d 1 2 − 2 1 2 2 3 − 3 2 1 1 3 1 − 1 3 2 and the cubic Casimir (q = 1) 6 ± m 1 Φ = X X X qX X X X2 + (d X + d (X + X )). PIV 1 2 3 − 2 1 3 − 2 1 1 q 1 1 2 2 3 − corresponds to the P IV case in the table of cubics 1. 6.3.2. NC PainlevéII. Consider the potential (6.83) Φ = X X X qX X X + (q 1)(X +Ω X + X ) PII 1 2 3 − 2 1 3 − 1 2 2 3 The algebraic relations corresponding to the Jacobian ideal are (6.84) X X qX X = (q 1); X X qX X = (q 1); X X qX X = (q 1)Ω 1 2 − 2 1 − 2 3 − 3 2 − 3 1 − 1 3 − 2 and the Poisson limit gives the Casimir cubics for the Miwa-Jimbo Painlevé II cases: φ = x x x + x + ω x + x cl,PII − 1 2 3 1 2 2 3 The authors of [3] argue that, in the framework of study of “orbifold singularities”, one should take the relation on the moduli space n n n n/2 w (6.85) x1x2x3 (e1 x1 + e2 x2 + e3 x3)+2(e1e2e3) Tn( 1/2 )=0, − −2(e1e2e3) QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 37 where Tn(w) = cos(n arccos w) is the n- th Chebyshev polynomial. iπ Taking e1 = e2 = e3 = exp( n ) and n = 1 (which means T1(w) = w) we have the expression x x x + x + x + x w =0, 1 2 3 1 2 3 − so the Miwa-Jimbo PainlevéCasimir cubic can be considered as the n = 1 member of the family (6.85).

6.4. Le Bryun -Witten algebras. Our ﬁnal remark is that the generalised Sklyanin- γ Painlev´ealgebra with potential Φ gives an example of the conformal sl2- enveloping algebra Uabc(sl2) ([26]). It corresponds to the choice of the parameters a = c = q; m = m = 1; e = e =0,e = 1; γ = b : 1 2 1 3 2 − − X1X2 qX2X1 = X3 (6.86) X X − qX X = X  2 3 3 2 1 X X − qX X = γX2 + X +1  3 1 − 1 3 − 2 2 This algebra corresponds to the generalised Sklyanin with β = γ = 0, a = b = c = q, case (2) of Theorem 4.10 The central element is 1+ q 1 Ω = (q2 1)X X X γ X3 + qX2 + X2 + qX2. LBW − 3 2 1 − q(1 + q + q2) 2 1 q 2 3

It was proved by Le Bruyn in ([26]) that the conformal sl2 enevloping algebras are Auslander regular and have the Cohen-Macaulay property as finitely generated (left) filtered rings. We observe now that, following the results of Artin, Tate and -Van den Bergh ([1]), one can construct a cubic divisor C ֒ P2 for any three dimensional Auslander-regular algebra and the algebra is defined→ by the divisor and an automorphism σ : C C. This divisor is defined by→ the equation γX qX X 2 − 1 3 (6.87) [C] = det X1 0 qX2 =0,  qX X − 0  − 3 2 where the determinant is calculated quantically as follows: γqX3 + (q3 1)X X X =0 − 2 − 1 2 3 and defines a conic ( γqX2 + (q3 1)X X = 0) and a line X = 0. The automor- − 2 − 1 3 2 phism σ is given on the line by σ(X1 : 0 : X3) = (X1 : 0 : qX3) and on the conic 1 3 by σ(X1 : X2 : X3) = (qX : X2 : q− X3). Thus, we see that if γ 0, q = 1 then ≡3 6 the divisor gives the triangular configuration X1X2X3 = 0 and if q = 1 then the divisor degeneratres in a triple line.

7. del Pezzo of degree 3 and open problems In this section, we summarise our results concerning del Pezzo of degree 3 and their quantisation in three tables and highlight some open problems for the future. Let us start by describing Table 4. The first column contains a list of double affine Hecke algebras. The elliptic (1) DAHA of type E6 is due to Rains, [44], while the GDAHA of type E6 is due to [16]. The abbreviation “Deg. GDAHA” corresponds to some Whittaker degenerations (1) ˇ of the E6 GDAHAe [12, 28], the CC1 DAHA is due to Cherednik [11, 47, 31], 38 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV while the abbreviation “Deg. DAHA” correspond to the algebras obtained in [27] by Whittaker degeneration. The second column is the polynomial φ such that φ is the center of the corresponding (elliptic or generalised or degenerate) DAHAM for q = 1: for the cases of Elliptic DAHA, this was conjectured in [16], for GDAHA it was proved in [16], for the CCˇ 1 in [32] and all other cases in [27, 28]. As discussed is Section 5, the projective completion φ is a del Pezzo of degree 3 with divisor D - this is specified in the third columnM of table 4. The table is split∞ vertically by a double line - the whole right side of the table is due to H. Sakai [48], and we have used his notation here. Before explaining what this double line represents, let us recall the definition of an Okamoto pair (X, ∆): this is a pair (X, ∆) where X is a generalised Halphen surface, namely the blow up of 9 points in P2 in non generic position, and ∆ is a divisor that tells us the position of such 9 points. Note that ∆ has the same configuration as a degenerate elliptic curve in the classification by Kodaira-Neron. In other words, the 9 non generic points lie at the intersection between ∆ and a generic elliptic curve in P2. The generalised Halphen surfaces are uniquely determined by their divisor ∆ listed in the fifth column. Some of these divisors have multiple points on them. Starting from the fifth row, at the intersection of lines we always have a multiple point, this is denoted by an empty circle. The order of this point can be calculated by removing from the number 9 the order of all other points. The single bullet points mean simple points, the bullets with a circle and a number next to them mean multiple points with the order specified by the number. In the last column we show the blow up of such divisor ∆ at the multiple points. Sakai labels the generalised Halphen surface according to the affine Weyl group (1) corresponding to the intersection matrix of the divisor. Note that in the case A0 Sakai uses two notations according to the divisor, no star means ∆ is a smooth elliptic curve, one star means ∆ is rational curve with a node. These labels are given in the fourth column. The first line of the table corresponds to the elliptic Painlevéequation, the next three lines to the multiplicative or q-difference Painlevéequations and the last eight rows correspond to the Painlevédifferential equations. There are also additive difference Painlevéequations, which we give in Table 6 because the corresponding quantum algebra is not Calabi-Yau [30]. Finally, there are also high dimension multiplicative difference Painlevéequations the quantum description of which is postponed [28]. In the case of the Painlevédifferential equations, the left and right sides of the table are related by the so-called Riemann-Hilbert correspondence - this was proved by several authors, a nice unified approach can be found in [54]. The basic idea is that the Okamoto pair corresponds to the space of initial conditions of the given equation, while the Looijenga pair corresponds to the monodromy manifold. Okamoto’s theory of initial value spaces [37], developed by Sakai [48], provides a beautiful unification of differential and discrete equations. Whether differential or discrete, initial values for any nonlinear equation, can be regular (meaning the solution will be analytic around the initial point) or can be unbounded (reflecting the existence of a singularity at the initial point). Okamoto compactified this space to the complex projective plane and showed that any subsequent indeterminacy can QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 39 be removed by resolving the base points through blowup techniques from algebraic geometry. It is a miraculous fact that nine blowups leads to a regularisation of the whole space for all differential and discrete Painlevéequations. For discrete Painlevé equations, there is no satisfactory concept of monodromy manifold - it is true that each additive discrete Painlevéequation comes from the Backlünd transformations of one of the differential ones, so that one could use the monodromy manifold associated to the latter, however without a direct isomonodromic approach, interesting dynamical behaviour may be lost. Moreover for the multiplicative discrete Painlevé equations, a notion of monodromy manifold is completely missing. This leads us to Conjecture 7.1. For the elliptic and multiplicative/additive discrete Painlevé equations, the Riemann Hilbert correspondence assigns to the generalised Halphen surface in Table 4 the corresponding Looijenga pair. Intuitively speaking, evidence for this conjecture is provided by the fact that the polynomials defining the divisors D in the first four lines of Table 4 are the same as those defining the corresponding∞ Halphen divisors ∆ - i.e. D = ∆ for the first four lines in the table. ∞ We list the quantum results in Table 5 and 6. All the quantum algebras in Table 5 are specialisations of the generalised Sklyanin-Painlevéalgebra introduced in subsection 4.8. We conclude by mentioning the relation between the quantum algebras in Table 5 and the matrix generalisations of the Painlevéequations. Building upon work by Retakh and the third author [45], in [5] a set of non-commutative relations which are non-commutative analogues of monodromy data relations for the Painlevé II equation was constructed. The interesting feature of these non-commutative relations is that by taking the scalar degeneration of the non-commutative operator q, one obtains our quantum PainlevéII monodromy variety. This observation opens the possibility of relating higher rank Elliptic/Generalised DAHA to the theory of matrix Painlevéequations. 40 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

φ Gen. Center del Pezzo Halphen DAHA for q = 1 Halphen Blow up divisor D∞ surface divisor ∆

Elliptic τx x x x3 x3 x3 (1) 1 2 3 + 1 + 2 + 3 A N.A. τx1x2x3+ 0 +a x2 + b x2 + c x2+ E6 1 1 1 2 1 3 3 3 3 +x1 + x2 + x3 +a2x1 + b2x2 + c2x3 + d 3 3 τx1x2x3 + x + x + e 1 2 (1)⋆ GDAHA τx x x + 2 2 2 1 2 3 A +a1x1 + b1x2 + c1x3+ 0 (1) x3 x3 N.A. E + 1 + 2 6 +a2x1 + b2x2 + c2x3 + d

3 τx1x2x3 + x 1 (1) 2 2 2 3 A1 +a1x1 + b1x2 + c1x3+ τx1x2x3 + x1 N.A. a x b x c x d Deg. + 2 1 + 2 2 + 2 3 + GDAHA (1) x1x2x3 A2 N.A. 2 2 2 x x x +a1x1 + b1x2 + c1x3+ 1 2 3

+a2x1 + b2x2 + c2x3 + d

x x x − x2 − x2 − x2+ DAHA 1 2 3 1 2 3 (1) x1x2x3 D4 • • • +ω x + ω x + ω x + ω 1 1 2 2 3 3 4 • • • CCˇ 1

x x x − x2 − x2+ (1) 1 2 3 1 2 D5 • • x1x2x3 ω1x1 + ω2x2 + ω3x3 + ω4 • • •

2 2 x1x2x3 − x − x + 1 2 (1) x1x2x3 D6 • • ω x + ω x + ω 1 1 2 2 4 • •

− 2 − 2 x x x (1) x1x2x3 x1 x2+ 1 2 3 D7 • 2 Deg. +ω x − x 1 1 2 •2 DAHA

(1) D8 • 4 − 2 − 2 − x x x x1x2x3 x1 x2 x2 1 2 3

x x x − x2 (1) 1 2 3 1+ E x1x2x3 6 • ω1x1 + ω2x2 + ω3x3 + ω4 • •2

2 (1) x x x − x + x1x2x3 1 2 3 1 E7 • − − +ω1x1 x2 1 •

(1) x1x2x3 − x1 − x2 + 1 x1x2x3 E8 • 9

Table 4. Results for del Pezzo of degree 3. QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 41

Central Polynomial φ Quantum relations potential element

3 3 3 x1x2x3 + x1 + x2 + x3 (4.31) ΦEG +ΨEG (4.33) a x2 b x2 c x2 + 1 1 + 1 2 + 1 3+ see (4.29), (4.30) +a2x1 + b2x2 + c2x3 + d 3 3 x1x2x3 + x1 + x2+ 2 2 2 (4.38) Φα,β,0 +ΨEG +a1x1 + b1x2 + c1x3+ (4.39)

+a2x1 + b2x2 + c2x3 + d with γ =0 see (4.37), (4.30)

3 x1x2x3 + x1 (4.38) Φα,0,0 +ΨEG (4.39) 2 2 2 +a1x1 + b1x2 + c1x3+ with β = γ =0 see (4.37), (4.30) with +a2x1 + b2x2 + c2x3 + d β =0

ΦEG +ΨEG x1x2x3 (4.31) (4.35) with t =0 +a x2 + b x2 + c x2+ 1 1 1 2 1 3 with t =0 see (4.29), (4.30) +a2x1 + b2x2 + c2x3 + d (d) φ , d = PVI,...,PI (d) (d) Φ P (1.5) with ǫ , Ω UZ (1.8) see (2.13) i i see (1.7) in (2.14) (2.15)

Table 5. Quantum counterpart of Table 4 (we have squashed the last eight lines of Table 4 into one).

Halphen Polynomial φ Quantum relations Divisor ∆ surface

∗∗ 3 2 x2 = x2 =0 (1) x x x3 1 2 A 1 − 2 0 x3x2 + x2x3 =0

x x + x x x2 =0 ∗ x2x x2x 2 3 3 2 1 (1) 2 3 1 2 2 − A1 − x2 = x2x1 + x1x2 =0

∗ 3 3 2 2 (1) x1 + x3 x1 = x2 =0 A2

Table 6. Non Calabi-Yau cases 42 LEONID CHEKHOV, MARTA MAZZOCCO, VLADIMIR RUBTSOV

References [1] M. Artin, J. Tate, and M. Van den Bergh. Modules over regular algebras of dimension 3. Invent. Math., 106(2):335–388, 1991. [2] Michael Artin and William F. Schelter. Graded algebras of global dimension 3. Adv. in Math., 66(2):171–216, 1987. [3] David Berenstein, Vishnu Jejjala, and Robert G. Leigh. Marginal and relevant deformations of N = 4 field theories and non-commutative moduli spaces of vacua. Nuclear Phys. B, 589(1-2):196–248, 2000. [4] Roland Berger and Rachel Taillefer. Poincaré-Birkhoff-Witt deformations of Calabi-Yau algebras. J. Noncommut. Geom., 1(2):241–270, 2007. [5] M. Bertola, M. Cafasso, and V. Rubtsov. Noncommutative Painlevéequations and systems of Calogero type. Comm. Math. Phys., 363(2):503–530, 2018. [6] Pierrick Bousseau. Quantum mirrors of log-calabi-yau surfaces and higher genus curve count- ing. arXiv:1808:07336 v1, 2018. [7] Alexander Braverman and Dennis Gaitsgory. Poincaré-Birkhoff-Witt theorem for quadratic algebras of Koszul type. J. Algebra, 181(2):315–328, 1996. [8] Tom Bridgeland and Ivan Smith. Quadratic differentials as stability conditions. Publ. Math. Inst. Hautes Etudes´ Sci., 121:155–278, 2015. [9] Leonid Chekhov and Marta Mazzocco. Shear coordinate description of the quantized versal unfolding of a D4 singularity. J. Phys. A, 43(44):442002, 13, 2010. [10] Leonid O. Chekhov, Marta Mazzocco, and Vladimir N. Rubtsov. Painlevémonodromy manifolds, decorated character varieties, and cluster algebras. Int. Math. Res. Not. IMRN, (24):7639–7691, 2017. [11] Ivan Cherednik. Double affine Hecke algebras, Knizhnik-Zamolodchikov equations, and Mac- donald’s operators. Internat. Math. Res. Notices, (9):171–180, 1992. [12] Ivan Cherednik. Whittaker limits of difference spherical functions. Int. Math. Res. Not. IMRN, (20):3793–3842, 2009. [13] F Denef. Les houches lecture in constructing string vacua. In String Theory and the Real World. From Particle Physics to Astrophysics, Proc. Summer School in Theor. Physics, pages 483–610. North Holland, 2008. [14] Vasiliy Dolgushev. The Van den Bergh duality and the modular symmetry of a Poisson variety. Selecta Math. (N.S.), 14(2):199–228, 2009. [15] Pavel Etingof and Victor Ginzburg. Noncommutative del Pezzo surfaces and Calabi-Yau algebras. J. Eur. Math. Soc. (JEMS), 12(6):1371–1416, 2010. [16] Pavel Etingof, Alexei Oblomkov, and Eric Rains. Generalized double affine Hecke algebras of rank 1 and quantized del Pezzo surfaces. Adv. Math., 212(2):749–796, 2007. [17] Hermann Flaschka and Alan C. Newell. Monodromy- and spectrum-preserving deformations. I. Comm. Math. Phys., 76(1):65–116, 1980. [18] Michael Gekhtman, Michael Shapiro, and Alek Vainshtein. Cluster algebras and Poisson geometry, volume 167 of Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, 2010. [19] B. R. Greene, M. R. Plesser, and S.-S. Roan. New constructions of mirror manifolds: probing moduli space far from Fermat points. In Essays on mirror manifolds, pages 408–448. Int. Press, Hong Kong, 1992. [20] Pierre-Paul Grivel. Une histoire du théorème de Poincaré-Birkhoff-Witt. Expo. Math., 22(2):145–184, 2004. [21] Mark Gross, Paul Hacking, and Sean Keel. Mirror symmetry for log Calabi-Yau surfaces I. Publ. Math. Inst. Hautes Etudes´ Sci., 122:65–168, 2015. [22] Natalia Iyudu and Stanislav Shkarin. Three dimensional Sklyanin algebras and Gröbner bases. J. Algebra, 470:379–419, 2017. [23] Natalia Iyudu and Stanislav Shkarin. Classification of quadratic and cubic pbw algebras on three generators. 2018. [24] Michio Jimbo and Tetsuji Miwa. Monodromy preserving deformation of linear ordinary differential equations with rational coefficients. II. Phys. D, 2(3):407–448, 1981. [25] Tom H. Koornwinder. The relationship between Zhedanov’s algebra AW(3) and the double affine Hecke algebra in the rank one case. SIGMA Symmetry Integrability Geom. Methods Appl., 3:Paper 063, 15, 2007. QUANTISED PAINLEVE´ MONODROMIES, SKLYANIN AND CY ALGEBRAS. 43

[26] Lieven Le Bruyn. Conformal sl2 enveloping algebras. Comm. Algebra, 23(4):1325–1362, 1995. [27] Marta Mazzocco. Confluences of the Painlevéequations, Cherednik algebras and q-Askey scheme. Nonlinearity, 29(9):2565–2608, 2016. [28] Marta Mazzocco. Whittaker degenerations of gdaha. in progress, 2019. [29] A. I. Molev and E. Ragoucy. Symmetries and invariants of twisted quantum algebras and associated Poisson algebras. Rev. Math. Phys., 20(2):173–198, 2008. [30] Izuru Mori and S. Paul Smith. The classification of 3-Calabi-Yau algebras with 3 generators and 3 quadratic relations. Math. Z., 287(1-2):215–241, 2017. [31] Masatoshi Noumi and Jasper V. Stokman. Askey-Wilson polynomials: an affine Hecke algebra approach. In Laredo Lectures on Orthogonal Polynomials and Special Functions, Adv. Theory Spec. Funct. Orthogonal Polynomials, pages 111–144. Nova Sci. Publ., Hauppauge, NY, 2004. [32] Alexei Oblomkov. Double affine Hecke algebras of rank 1 and affine cubic surfaces. Int. Math. Res. Not., (18):877–912, 2004. [33] Aleksandr V. Odesski. Rational degeneration of elliptic quadratic algebras. In Infinite anal- ysis, Part A, B (Kyoto, 1991), volume 16 of Adv. Ser. Math. Phys., pages 773–779. World Sci. Publ., River Edge, NJ, 1992. [34] A. V. Odesski˘ı. An analogue of the Sklyanin algebra. Funktsional. Anal. i Prilozhen., 20(2):78–79, 1986. [35] A. V. Odesski˘ıand B. L. Fe˘ıgin. Sklyanin’s elliptic algebras. The case of a point of finite order. Funktsional. Anal. i Prilozhen., 29(2):9–21, 95, 1995. [36] A. V. Odesski˘ıand V. N. Rubtsov. Polynomial Poisson algebras with a regular structure of symplectic leaves. Teoret. Mat. Fiz., 133(1):3–23, 2002. [37] Kazuo Okamoto. Sur les feuilletages associés aux équations du second ordre àpoints critiques fixes de P. Painlevé. Japan. J. Math. (N.S.), 5(1):1–79, 1979. [38] Giovanni Ortenzi, Vladimir Rubtsov, and Serge Roméo Tagne Pelap. On the Heisenberg invariance and the elliptic Poisson tensors. Lett. Math. Phys., 96(1-3):263–284, 2011. [39] P. Painlevé. Mémoire sur les équations différentielles dont l’intégrale générale est uniforme. Bull. Soc. Math. France, 28:201–261, 1900. [40] A. Polishchuk. Poisson structures and birational morphisms associated with bundles on elliptic curves. Internat. Math. Res. Notices, (13):683–703, 1998. [41] Alexander Polishchuk and Leonid Positselski. Quadratic algebras, volume 37 of University Lecture Series. American Mathematical Society, Providence, RI, 2005. [42] Stewart B. Priddy. Koszul resolutions. Trans. Amer. Math. Soc., 152:39–60, 1970. [43] E. Rains. Computer calculation. [44] E. Rains. Elliptic double affine hecke algebras. arXiv:1709.02989, 2017. [45] Vladimir Retakh and Vladimir Rubtsov. Noncommutative Toda chains, Hankel quasideter- minants and the PainlevéII equation. J. Phys. A, 43(50):505204, 13, 2010. [46] Shi-shyr Roan. Mirror symmetry of elliptic curves and Ising model. J. Geom. Phys., 20(2- 3):273–296, 1996. [47] Siddhartha Sahi. Nonsymmetric Koornwinder polynomials and duality. Ann. of Math. (2), 150(1):267–282, 1999. [48] Hidetaka Sakai. Rational surfaces associated with affine root systems and geometry of the Painlevéequations. Comm. Math. Phys., 220(1):165–229, 2001. [49] Boris Shoikhet. The PBW property for associative algebras as an integrability condition. Math. Res. Lett., 21(6):1407–1434, 2014. [50] S. Paul Smith. “Degenerate” 3-dimensional Sklyanin algebras are monomial algebras. J. Al- gebra, 358:74–86, 2012. [51] Andrew Strominger, Shing-Tung Yau, and Eric Zaslow. Mirror symmetry is T -duality. Nuclear Phys. B, 479(1-2):243–259, 1996. ∨ [52] Paul Terwilliger. The universal Askey-Wilson algebra and DAHA of type (C1 , C1). SIGMA Symmetry Integrability Geom. Methods Appl., 9:Paper 047, 40, 2013. [53] Ginzburg V.A. Calabi-yau algebras. 2006. [54] Marius van der Put and Masa-Hiko Saito. Moduli spaces for linear differential equations and the Painlevéequations. Ann. Inst. Fourier (Grenoble), 59(7):2611–2667, 2009. [55] Luc Vinet and Alexei Zhedanov. Quasi-linear algebras and integrability (the Heisenberg pic- ture). SIGMA Symmetry Integrability Geom. Methods Appl., 4:Paper 015, 22, 2008. [56] Chelsea Walton. Representation theory of three-dimensional Sklyanin algebras. Nuclear Phys. B, 860(1):167–185, 2012.